36 skills found · Page 1 of 2
meta-llama / Llama CookbookWelcome to the Llama Cookbook! This is your go to guide for Building with Llama: Getting started with Inference, Fine-Tuning, RAG. We also show you how to solve end to end problems using Llama model family and using them on various provider services
bentoml / BentoMLThe easiest way to serve AI apps and models - Build Model Inference APIs, Job queues, LLM apps, Multi-model pipelines, and more!
solidglue / Recommender System推荐系统入门指南，全面介绍了工业级推荐系统的理论知识（王树森推荐系统公开课-基于小红书的场景讲解工业界真实的推荐系统），如何基于TensorFlow2训练模型，如何实现高性能、高并发、高可用的Golang推理微服务。Comprehensively introduced the theory of industrial recommender system, how to trainning models based on TensorFlow2, how to implement the high-performance、high-concurrency and high-available inference services base on Golang.
Lightning-Universe / Stable Diffusion DeployLearn to serve Stable Diffusion models on cloud infrastructure at scale. This Lightning App shows load-balancing, orchestrating, pre-provisioning, dynamic batching, GPU-inference, micro-services working together via the Lightning Apps framework.
chrisneagu / FTC Skystone Dark Angels Romania 2020NOTICE This repository contains the public FTC SDK for the SKYSTONE (2019-2020) competition season. If you are looking for the current season's FTC SDK software, please visit the new and permanent home of the public FTC SDK: FtcRobotController repository Welcome! This GitHub repository contains the source code that is used to build an Android app to control a FIRST Tech Challenge competition robot. To use this SDK, download/clone the entire project to your local computer. Getting Started If you are new to robotics or new to the FIRST Tech Challenge, then you should consider reviewing the FTC Blocks Tutorial to get familiar with how to use the control system: FTC Blocks Online Tutorial Even if you are an advanced Java programmer, it is helpful to start with the FTC Blocks tutorial, and then migrate to the OnBot Java Tool or to Android Studio afterwards. Downloading the Project If you are an Android Studio programmer, there are several ways to download this repo. Note that if you use the Blocks or OnBot Java Tool to program your robot, then you do not need to download this repository. If you are a git user, you can clone the most current version of the repository: git clone https://github.com/FIRST-Tech-Challenge/SKYSTONE.git Or, if you prefer, you can use the "Download Zip" button available through the main repository page. Downloading the project as a .ZIP file will keep the size of the download manageable. You can also download the project folder (as a .zip or .tar.gz archive file) from the Downloads subsection of the Releases page for this repository. Once you have downloaded and uncompressed (if needed) your folder, you can use Android Studio to import the folder ("Import project (Eclipse ADT, Gradle, etc.)"). Getting Help User Documentation and Tutorials FIRST maintains online documentation with information and tutorials on how to use the FIRST Tech Challenge software and robot control system. You can access this documentation using the following link: SKYSTONE Online Documentation Note that the online documentation is an "evergreen" document that is constantly being updated and edited. It contains the most current information about the FIRST Tech Challenge software and control system. Javadoc Reference Material The Javadoc reference documentation for the FTC SDK is now available online. Click on the following link to view the FTC SDK Javadoc documentation as a live website: FTC Javadoc Documentation Documentation for the FTC SDK is also included with this repository. There is a subfolder called "doc" which contains several subfolders: The folder "apk" contains the .apk files for the FTC Driver Station and FTC Robot Controller apps. The folder "javadoc" contains the JavaDoc user documentation for the FTC SDK. Online User Forum For technical questions regarding the Control System or the FTC SDK, please visit the FTC Technology forum: FTC Technology Forum Release Information Version 5.5 (20200824-090813) Version 5.5 requires Android Studio 4.0 or later. New features Adds support for calling custom Java classes from Blocks OpModes (fixes SkyStone issue #161). Classes must be in the org.firstinspires.ftc.teamcode package. Methods must be public static and have no more than 21 parameters. Parameters declared as OpMode, LinearOpMode, Telemetry, and HardwareMap are supported and the argument is provided automatically, regardless of the order of the parameters. On the block, the sockets for those parameters are automatically filled in. Parameters declared as char or java.lang.Character will accept any block that returns text and will only use the first character in the text. Parameters declared as boolean or java.lang.Boolean will accept any block that returns boolean. Parameters declared as byte, java.lang.Byte, short, java.lang.Short, int, java.lang.Integer, long, or java.lang.Long, will accept any block that returns a number and will round that value to the nearest whole number. Parameters declared as float, java.lang.Float, double, java.lang.Double will accept any block that returns a number. Adds telemetry API method for setting display format Classic Monospace HTML (certain tags only) Adds blocks support for switching cameras. Adds Blocks support for TensorFlow Object Detection with a custom model. Adds support for uploading a custom TensorFlow Object Detection model in the Manage page, which is especially useful for Blocks and OnBotJava users. Shows new Control Hub blink codes when the WiFi band is switched using the Control Hub's button (only possible on Control Hub OS 1.1.2) Adds new warnings which can be disabled in the Advanced RC Settings Mismatched app versions warning Unnecessary 2.4 GHz WiFi usage warning REV Hub is running outdated firmware (older than version 1.8.2) Adds support for Sony PS4 gamepad, and reworks how gamepads work on the Driver Station Removes preference which sets gamepad type based on driver position. Replaced with menu which allows specifying type for gamepads with unknown VID and PID Attempts to auto-detect gamepad type based on USB VID and PID If gamepad VID and PID is not known, use type specified by user for that VID and PID If gamepad VID and PID is not known AND the user has not specified a type for that VID and PID, an educated guess is made about how to map the gamepad Driver Station will now attempt to automatically recover from a gamepad disconnecting, and re-assign it to the position it was assigned to when it dropped If only one gamepad is assigned and it drops: it can be recovered If two gamepads are assigned, and have different VID/PID signatures, and only one drops: it will be recovered If two gamepads are assigned, and have different VID/PID signatures, and BOTH drop: both will be recovered If two gamepads are assigned, and have the same VID/PID signatures, and only one drops: it will be recovered If two gamepads are assigned, and have the same VID/PID signatures, and BOTH drop: neither will be recovered, because of the ambiguity of the gamepads when they re-appear on the USB bus. There is currently one known edge case: if there are two gamepads with the same VID/PID signature plugged in, but only one is assigned, and they BOTH drop, it's a 50-50 chance of which one will be chosen for automatic recovery to the assigned position: it is determined by whichever one is re-enumerated first by the USB bus controller. Adds landscape user interface to Driver Station New feature: practice timer with audio cues New feature (Control Hub only): wireless network connection strength indicator (0-5 bars) New feature (Control Hub only): tapping on the ping/channel display will switch to an alternate display showing radio RX dBm and link speed (tap again to switch back) The layout will NOT autorotate. You can switch the layout from the Driver Station's settings menu. Breaking changes Removes support for Android versions 4.4 through 5.1 (KitKat and Lollipop). The minSdkVersion is now 23. Removes the deprecated LinearOpMode methods waitOneFullHardwareCycle() and waitForNextHardwareCycle() Enhancements Handles RS485 address of Control Hub automatically The Control Hub is automatically given a reserved address Existing configuration files will continue to work All addresses in the range of 1-10 are still available for Expansion Hubs The Control Hub light will now normally be solid green, without blinking to indicate the address The Control Hub will not be shown on the Expansion Hub Address Change settings page Improves REV Hub firmware updater The user can now choose between all available firmware update files Version 1.8.2 of the REV Hub firmware is bundled into the Robot Controller app. Text was added to clarify that Expansion Hubs can only be updated via USB. Firmware update speed was reduced to improve reliability Allows REV Hub firmware to be updated directly from the Manage webpage Improves log viewer on Robot Controller Horizontal scrolling support (no longer word wrapped) Supports pinch-to-zoom Uses a monospaced font Error messages are highlighted New color scheme Attempts to force-stop a runaway/stuck OpMode without restarting the entire app Not all types of runaway conditions are stoppable, but if the user code attempts to talk to hardware during the runaway, the system should be able to capture it. Makes various tweaks to the Self Inspect screen Renames "OS version" entry to "Android version" Renames "WiFi Direct Name" to "WiFi Name" Adds Control Hub OS version, when viewing the report of a Control Hub Hides the airplane mode entry, when viewing the report of a Control Hub Removes check for ZTE Speed Channel Changer Shows firmware version for all Expansion and Control Hubs Reworks network settings portion of Manage page All network settings are now applied with a single click The WiFi Direct channel of phone-based Robot Controllers can now be changed from the Manage page WiFi channels are filtered by band (2.4 vs 5 GHz) and whether they overlap with other channels The current WiFi channel is pre-selected on phone-based Robot Controllers, and Control Hubs running OS 1.1.2 or later. On Control Hubs running OS 1.1.2 or later, you can choose to have the system automatically select a channel on the 5 GHz band Improves OnBotJava New light and dark themes replace the old themes (chaos, github, chrome,...) the new default theme is light and will be used when you first update to this version OnBotJava now has a tabbed editor Read-only offline mode Improves function of "exit" menu item on Robot Controller and Driver Station Now guaranteed to be fully stopped and unloaded from memory Shows a warning message if a LinearOpMode exists prematurely due to failure to monitor for the start condition Improves error message shown when the Driver Station and Robot Controller are incompatible with each other Driver Station OpMode Control Panel now disabled while a Restart Robot is in progress Disables advanced settings related to WiFi direct when the Robot Controller is a Control Hub. Tint phone battery icons on Driver Station when low/critical. Uses names "Control Hub Portal" and "Control Hub" (when appropriate) in new configuration files Improve I2C read performance Very large improvement on Control Hub; up to ~2x faster with small (e.g. 6 byte) reads Not as apparent on Expansion Hubs connected to a phone Update/refresh build infrastructure Update to 'androidx' support library from 'com.android.support:appcompat', which is end-of-life Update targetSdkVersion and compileSdkVersion to 28 Update Android Studio's Android plugin to latest Fix reported build timestamp in 'About' screen Add sample illustrating manual webcam use: ConceptWebcam Bug fixes Fixes SkyStone issue #248 Fixes SkyStone issue #232 and modifies bulk caching semantics to allow for cache-preserving MANUAL/AUTO transitions. Improves performance when REV 2M distance sensor is unplugged Improves readability of Toast messages on certain devices Allows a Driver Station to connect to a Robot Controller after another has disconnected Improves generation of fake serial numbers for UVC cameras which do not provide a real serial number Previously some devices would assign such cameras a serial of 0:0 and fail to open and start streaming Fixes ftc_app issue #638. Fixes a slew of bugs with the Vuforia camera monitor including: Fixes bug where preview could be displayed with a wonky aspect ratio Fixes bug where preview could be cut off in landscape Fixes bug where preview got totally messed up when rotating phone Fixes bug where crosshair could drift off target when using webcams Fixes issue in UVC driver on some devices (ftc_app 681) if streaming was started/stopped multiple times in a row Issue manifested as kernel panic on devices which do not have this kernel patch. On affected devices which do have the patch, the issue was manifest as simply a failure to start streaming. The Tech Team believes that the root cause of the issue is a bug in the Linux kernel XHCI driver. A workaround was implemented in the SDK UVC driver. Fixes bug in UVC driver where often half the frames from the camera would be dropped (e.g. only 15FPS delivered during a streaming session configured for 30FPS). Fixes issue where TensorFlow Object Detection would show results whose confidence was lower than the minimum confidence parameter. Fixes a potential exploitation issue of CVE-2019-11358 in OnBotJava Fixes changing the address of an Expansion Hub with additional Expansion Hubs connected to it Preserves the Control Hub's network connection when "Restart Robot" is selected Fixes issue where device scans would fail while the Robot was restarting Fix RenderScript usage Use androidx.renderscript variant: increased compatibility Use RenderScript in Java mode, not native: simplifies build Fixes webcam-frame-to-bitmap conversion problem: alpha channel wasn't being initialized, only R, G, & B Fixes possible arithmetic overflow in Deadline Fixes deadlock in Vuforia webcam support which could cause 5-second delays when stopping OpMode Version 5.4 (20200108-101156) Fixes SkyStone issue #88 Adds an inspection item that notes when a robot controller (Control Hub) is using the factory default password. Fixes SkyStone issue #61 Fixes SkyStone issue #142 Fixes ftc_app issue #417 by adding more current and voltage monitoring capabilities for REV Hubs. Fixes a crash sometimes caused by OnBotJava activity Improves OnBotJava autosave functionality ftc_app #738 Fixes system responsiveness issue when an Expansion Hub is disconnected Fixes issue where IMU initialization could prevent Op Modes from stopping Fixes issue where AndroidTextToSpeech.speak() would fail if it was called too early Adds telemetry.speak() methods and blocks, which cause the Driver Station (if also updated) to speak text Adds and improves Expansion Hub-related warnings Improves Expansion Hub low battery warning Displays the warning immediately after the hub reports it Specifies whether the condition is current or occurred temporarily during an OpMode run Displays which hubs reported low battery Displays warning when hub loses and regains power during an OpMode run Fixes the hub's LED pattern after this condition Displays warning when Expansion Hub is not responding to commands Specifies whether the condition is current or occurred temporarily during an OpMode run Clarifies warning when Expansion Hub is not present at startup Specifies that this condition requires a Robot Restart before the hub can be used. The hub light will now accurately reflect this state Improves logging and reduces log spam during these conditions Syncs the Control Hub time and timezone to a connected web browser programming the robot, if a Driver Station is not available. Adds bulk read functionality for REV Hubs A bulk caching mode must be set at the Hub level with LynxModule#setBulkCachingMode(). This applies to all relevant SDK hardware classes that reference that Hub. The following following Hub bulk caching modes are available: BulkCachingMode.OFF (default): All hardware calls operate as usual. Bulk data can read through LynxModule#getBulkData() and processed manually. BulkCachingMode.AUTO: Applicable hardware calls are served from a bulk read cache that is cleared/refreshed automatically to ensure identical commands don't hit the same cache. The cache can also be cleared manually with LynxModule#clearBulkCache(), although this is not recommended. (advanced users) BulkCachingMode.MANUAL: Same as BulkCachingMode.AUTO except the cache is never cleared automatically. To avoid getting stale data, the cache must be manually cleared at the beginning of each loop body or as the user deems appropriate. Removes PIDF Annotation values added in Rev 5.3 (to AndyMark, goBILDA and TETRIX motor configurations). The new motor types will still be available but their Default control behavior will revert back to Rev 5.2 Adds new ConceptMotorBulkRead sample Opmode to demonstrate and compare Motor Bulk-Read modes for reducing I/O latencies. Version 5.3 (20191004-112306) Fixes external USB/UVC webcam support Makes various bugfixes and improvements to Blocks page, including but not limited to: Many visual tweaks Browser zoom and window resize behave better Resizing the Java preview pane works better and more consistently across browsers The Java preview pane consistently gets scrollbars when needed The Java preview pane is hidden by default on phones Internet Explorer 11 should work Large dropdown lists display properly on lower res screens Disabled buttons are now visually identifiable as disabled A warning is shown if a user selects a TFOD sample, but their device is not compatible Warning messages in a Blocks op mode are now visible by default. Adds goBILDA 5201 and 5202 motors to Robot Configurator Adds PIDF Annotation values to AndyMark, goBILDA and TETRIX motor configurations. This has the effect of causing the RUN_USING_ENCODERS and RUN_TO_POSITION modes to use PIDF vs PID closed loop control on these motors. This should provide more responsive, yet stable, speed control. PIDF adds Feedforward control to the basic PID control loop. Feedforward is useful when controlling a motor's speed because it "anticipates" how much the control voltage must change to achieve a new speed set-point, rather than requiring the integrated error to change sufficiently. The PIDF values were chosen to provide responsive, yet stable, speed control on a lightly loaded motor. The more heavily a motor is loaded (drag or friction), the more noticable the PIDF improvement will be. Fixes startup crash on Android 10 Fixes ftc_app issue #712 (thanks to FROGbots-4634) Fixes ftc_app issue #542 Allows "A" and lowercase letters when naming device through RC and DS apps. Version 5.2 (20190905-083277) Fixes extra-wide margins on settings activities, and placement of the new configuration button Adds Skystone Vuforia image target data. Includes sample Skystone Vuforia Navigation op modes (Java). Includes sample Skystone Vuforia Navigation op modes (Blocks). Adds TensorFlow inference model (.tflite) for Skystone game elements. Includes sample Skystone TensorFlow op modes (Java). Includes sample Skystone TensorFlow op modes (Blocks). Removes older (season-specific) sample op modes. Includes 64-bit support (to comply with Google Play requirements). Protects against Stuck OpModes when a Restart Robot is requested. (Thanks to FROGbots-4634) (ftc_app issue #709) Blocks related changes: Fixes bug with blocks generated code when hardware device name is a java or javascript reserved word. Shows generated java code for blocks, even when hardware items are missing from the active configuration. Displays warning icon when outdated Vuforia and TensorFlow blocks are used (SkyStone issue #27) Version 5.1 (20190820-222104) Defines default PIDF parameters for the following motors: REV Core Hex Motor REV 20:1 HD Hex Motor REV 40:1 HD Hex Motor Adds back button when running on a device without a system back button (such as a Control Hub) Allows a REV Control Hub to update the firmware on a REV Expansion Hub via USB Fixes SkyStone issue #9 Fixes ftc_app issue #715 Prevents extra DS User clicks by filtering based on current state. Prevents incorrect DS UI state changes when receiving new OpMode list from RC Adds support for REV Color Sensor V3 Adds a manual-refresh DS Camera Stream for remotely viewing RC camera frames. To show the stream on the DS, initialize but do not run a stream-enabled opmode, select the Camera Stream option in the DS menu, and tap the image to refresh. This feature is automatically enabled when using Vuforia or TFOD—no additional RC configuration is required for typical use cases. To hide the stream, select the same menu item again. Note that gamepads are disabled and the selected opmode cannot be started while the stream is open as a safety precaution. To use custom streams, consult the API docs for CameraStreamServer#setSource and CameraStreamSource. Adds many Star Wars sounds to RobotController resources. Added SKYSTONE Sounds Chooser Sample Program. Switches out startup, connect chimes, and error/warning sounds for Star Wars sounds Updates OnBot Java to use a WebSocket for communication with the robot The OnBot Java page no longer has to do a full refresh when a user switches from editing one file to another Known issues: Camera Stream The Vuforia camera stream inherits the issues present in the phone preview (namely ftc_app issue #574). This problem does not affect the TFOD camera stream even though it receives frames from Vuforia. The orientation of the stream frames may not always match the phone preview. For now, these frames may be rotated manually via a custom CameraStreamSource if desired. OnBotJava Browser back button may not always work correctly It's possible for a build to be queued, but not started. The OnBot Java build console will display a warning if this occurs. A user might not realize they are editing a different file if the user inadvertently switches from one file to another since this switch is now seamless. The name of the currently open file is displayed in the browser tab. Version 5.0 (built on 19.06.14) Support for the REV Robotics Control Hub. Adds a Java preview pane to the Blocks editor. Adds a new offline export feature to the Blocks editor. Display wifi channel in Network circle on Driver Station. Adds calibration for Logitech C270 Updates build tooling and target SDK. Compliance with Google's permissions infrastructure (Required after build tooling update). Keep Alives to mitigate the Motorola wifi scanning problem. Telemetry substitute no longer necessary. Improves Vuforia error reporting. Fixes ftctechnh/ftc_app issues 621, 713. Miscellaneous bug fixes and improvements. Version 4.3 (built on 18.10.31) Includes missing TensorFlow-related libraries and files. Version 4.2 (built on 18.10.30) Includes fix to avoid deadlock situation with WatchdogMonitor which could result in USB communication errors. Comm error appeared to require that user disconnect USB cable and restart the Robot Controller app to recover. robotControllerLog.txt would have error messages that included the words "E RobotCore: lynx xmit lock: #### abandoning lock:" Includes fix to correctly list the parent module address for a REV Robotics Expansion Hub in a configuration (.xml) file. Bug in versions 4.0 and 4.1 would incorrect list the address module for a parent REV Robotics device as "1". If the parent module had a higher address value than the daisy-chained module, then this bug would prevent the Robot Controller from communicating with the downstream Expansion Hub. Added requirement for ACCESS_COARSE_LOCATION to allow a Driver Station running Android Oreo to scan for Wi-Fi Direct devices. Added google() repo to build.gradle because aapt2 must be downloaded from the google() repository beginning with version 3.2 of the Android Gradle Plugin. Important Note: Android Studio users will need to be connected to the Internet the first time build the ftc_app project. Internet connectivity is required for the first build so the appropriate files can be downloaded from the Google repository. Users should not need to be connected to the Internet for subsequent builds. This should also fix buid issue where Android Studio would complain that it "Could not find com.android.tools.lint:lint-gradle:26.1.4" (or similar). Added support for REV Spark Mini motor controller as part of the configuration menu for a servo/PWM port on the REV Expansion Hub. Provide examples for playing audio files in an Op Mode. Block Development Tool Changes Includes a fix for a problem with the Velocity blocks that were reported in the FTC Technology forum (Blocks Programming subforum). Change the "Save completed successfully." message to a white color so it will contrast with a green background. Fixed the "Download image" feature so it will work if there are text blocks in the op mode. Introduce support for Google's TensorFlow Lite technology for object detetion for 2018-2019 game. TensorFlow lite can recognize Gold Mineral and Silver Mineral from 2018-2019 game. Example Java and Block op modes are included to show how to determine the relative position of the gold block (left, center, right). Version 4.1 (released on 18.09.24) Changes include: Fix to prevent crash when deprecated configuration annotations are used. Change to allow FTC Robot Controller APK to be auto-updated using FIRST Global Control Hub update scripts. Removed samples for non supported / non legal hardware. Improvements to Telemetry.addData block with "text" socket. Updated Blocks sample op mode list to include Rover Ruckus Vuforia example. Update SDK library version number. Version 4.0 (released on 18.09.12) Changes include: Initial support for UVC compatible cameras If UVC camera has a unique serial number, RC will detect and enumerate by serial number. If UVC camera lacks a unique serial number, RC will only support one camera of that type connected. Calibration settings for a few cameras are included (see TeamCode/src/main/res/xml/teamwebcamcalibrations.xml for details). User can upload calibration files from Program and Manage web interface. UVC cameras seem to draw a fair amount of electrical current from the USB bus. This does not appear to present any problems for the REV Robotics Control Hub. This does seem to create stability problems when using some cameras with an Android phone-based Robot Controller. FTC Tech Team is investigating options to mitigate this issue with the phone-based Robot Controllers. Updated sample Vuforia Navigation and VuMark Op Modes to demonstrate how to use an internal phone-based camera and an external UVC webcam. Support for improved motor control. REV Robotics Expansion Hub firmware 1.8 and greater will support a feed forward mechanism for closed loop motor control. FTC SDK has been modified to support PIDF coefficients (proportional, integral, derivative, and feed forward). FTC Blocks development tool modified to include PIDF programming blocks. Deprecated older PID-related methods and variables. REV's 1.8.x PIDF-related changes provide a more linear and accurate way to control a motor. Wireless Added 5GHz support for wireless channel changing for those devices that support it. Tested with Moto G5 and E4 phones. Also tested with other (currently non-approved) phones such as Samsung Galaxy S8. Improved Expansion Hub firmware update support in Robot Controller app Changes to make the system more robust during the firmware update process (when performed through Robot Controller app). User no longer has to disconnect a downstream daisy-chained Expansion Hub when updating an Expansion Hub's firmware. If user is updating an Expansion Hub's firmware through a USB connection, he/she does not have to disconnect RS485 connection to other Expansion Hubs. The user still must use a USB connection to update an Expansion Hub's firmware. The user cannot update the Expansion Hub firmware for a downstream device that is daisy chained through an RS485 connection. If an Expansion Hub accidentally gets "bricked" the Robot Controller app is now more likely to recognize the Hub when it scans the USB bus. Robot Controller app should be able to detect an Expansion Hub, even if it accidentally was bricked in a previous update attempt. Robot Controller app should be able to install the firmware onto the Hub, even if if accidentally was bricked in a previous update attempt. Resiliency FTC software can detect and enable an FTDI reset feature that is available with REV Robotics v1.8 Expansion Hub firmware and greater. When enabled, the Expansion Hub can detect if it hasn't communicated with the Robot Controller over the FTDI (USB) connection. If the Hub hasn't heard from the Robot Controller in a while, it will reset the FTDI connection. This action helps system recover from some ESD-induced disruptions. Various fixes to improve reliability of FTC software. Blocks Fixed errors with string and list indices in blocks export to java. Support for USB connected UVC webcams. Refactored optimized Blocks Vuforia code to support Rover Ruckus image targets. Added programming blocks to support PIDF (proportional, integral, derivative and feed forward) motor control. Added formatting options (under Telemetry and Miscellaneous categories) so user can set how many decimal places to display a numerical value. Support to play audio files (which are uploaded through Blocks web interface) on Driver Station in addition to the Robot Controller. Fixed bug with Download Image of Blocks feature. Support for REV Robotics Blinkin LED Controller. Support for REV Robotics 2m Distance Sensor. Added support for a REV Touch Sensor (no longer have to configure as a generic digital device). Added blocks for DcMotorEx methods. These are enhanced methods that you can use when supported by the motor controller hardware. The REV Robotics Expansion Hub supports these enhanced methods. Enhanced methods include methods to get/set motor velocity (in encoder pulses per second), get/set PIDF coefficients, etc.. Modest Improvements in Logging Decrease frequency of battery checker voltage statements. Removed non-FTC related log statements (wherever possible). Introduced a "Match Logging" feature. Under "Settings" a user can enable/disable this feature (it's disabled by default). If enabled, user provides a "Match Number" through the Driver Station user interface (top of the screen). The Match Number is used to create a log file specifically with log statements from that particular Op Mode run. Match log files are stored in /sdcard/FIRST/matlogs on the Robot Controller. Once an op mode run is complete, the Match Number is cleared. This is a convenient way to create a separate match log with statements only related to a specific op mode run. New Devices Support for REV Robotics Blinkin LED Controller. Support for REV Robotics 2m Distance Sensor. Added configuration option for REV 20:1 HD Hex Motor. Added support for a REV Touch Sensor (no longer have to configure as a generic digital device). Miscellaneous Fixed some errors in the definitions for acceleration and velocity in our javadoc documentation. Added ability to play audio files on Driver Station When user is configuring an Expansion Hub, the LED on the Expansion Hub will change blink pattern (purple-cyan) to indicate which Hub is currently being configured. Renamed I2cSensorType to I2cDeviceType. Added an external sample Op Mode that demonstrates localization using 2018-2019 (Rover Ruckus presented by QualComm) Vuforia targets. Added an external sample Op Mode that demonstrates how to use the REV Robotics 2m Laser Distance Sensor. Added an external sample Op Mode that demonstrates how to use the REV Robotics Blinkin LED Controller. Re-categorized external Java sample Op Modes to "TeleOp" instead of "Autonomous". Known issues: Initial support for UVC compatible cameras UVC cameras seem to draw significant amount of current from the USB bus. This does not appear to present any problems for the REV Robotics Control Hub. This does seem to create stability problems when using some cameras with an Android phone-based Robot Controller. FTC Tech Team is investigating options to mitigate this issue with the phone-based Robot Controllers. There might be a possible deadlock which causes the RC to become unresponsive when using a UVC webcam with a Nougat Android Robot Controller. Wireless When user selects a wireless channel, this channel does not necessarily persist if the phone is power cycled. Tech Team is hoping to eventually address this issue in a future release. Issue has been present since apps were introduced (i.e., it is not new with the v4.0 release). Wireless channel is not currently displayed for WiFi Direct connections. Miscellaneous The blink indication feature that shows which Expansion Hub is currently being configured does not work for a newly created configuration file. User has to first save a newly created configuration file and then close and re-edit the file in order for blink indicator to work. Version 3.6 (built on 17.12.18) Changes include: Blocks Changes Uses updated Google Blockly software to allow users to edit their op modes on Apple iOS devices (including iPad and iPhone). Improvement in Blocks tool to handle corrupt op mode files. Autonomous op modes should no longer get switched back to tele-op after re-opening them to be edited. The system can now detect type mismatches during runtime and alert the user with a message on the Driver Station. Updated javadoc documentation for setPower() method to reflect correct range of values (-1 to +1). Modified VuforiaLocalizerImpl to allow for user rendering of frames Added a user-overrideable onRenderFrame() method which gets called by the class's renderFrame() method. Version 3.5 (built on 17.10.30) Changes with version 3.5 include: Introduced a fix to prevent random op mode stops, which can occur after the Robot Controller app has been paused and then resumed (for example, when a user temporarily turns off the display of the Robot Controller phone, and then turns the screen back on). Introduced a fix to prevent random op mode stops, which were previously caused by random peer disconnect events on the Driver Station. Fixes issue where log files would be closed on pause of the RC or DS, but not re-opened upon resume. Fixes issue with battery handler (voltage) start/stop race. Fixes issue where Android Studio generated op modes would disappear from available list in certain situations. Fixes problem where OnBot Java would not build on REV Robotics Control Hub. Fixes problem where OnBot Java would not build if the date and time on the Robot Controller device was "rewound" (set to an earlier date/time). Improved error message on OnBot Java that occurs when renaming a file fails. Removed unneeded resources from android.jar binaries used by OnBot Java to reduce final size of Robot Controller app. Added MR_ANALOG_TOUCH_SENSOR block to Blocks Programming Tool. Version 3.4 (built on 17.09.06) Changes with version 3.4 include: Added telemetry.update() statement for BlankLinearOpMode template. Renamed sample Block op modes to be more consistent with Java samples. Added some additional sample Block op modes. Reworded OnBot Java readme slightly. Version 3.3 (built on 17.09.04) This version of the software includes improves for the FTC Blocks Programming Tool and the OnBot Java Programming Tool. Changes with verion 3.3 include: Android Studio ftc_app project has been updated to use Gradle Plugin 2.3.3. Android Studio ftc_app project is already using gradle 3.5 distribution. Robot Controller log has been renamed to /sdcard/RobotControllerLog.txt (note that this change was actually introduced w/ v3.2). Improvements in I2C reliability. Optimized I2C read for REV Expansion Hub, with v1.7 firmware or greater. Updated all external/samples (available through OnBot and in Android project folder). Vuforia Added support for VuMarks that will be used for the 2017-2018 season game. Blocks Update to latest Google Blockly release. Sample op modes can be selected as a template when creating new op mode. Fixed bug where the blocks would disappear temporarily when mouse button is held down. Added blocks for Range.clip and Range.scale. User can now disable/enable Block op modes. Fix to prevent occasional Blocks deadlock. OnBot Java Significant improvements with autocomplete function for OnBot Java editor. Sample op modes can be selected as a template when creating new op mode. Fixes and changes to complete hardware setup feature. Updated (and more useful) onBot welcome message. Known issues: Android Studio After updating to the new v3.3 Android Studio project folder, if you get error messages indicating "InvalidVirtualFileAccessException" then you might need to do a File->Invalidate Caches / Restart to clear the error. OnBot Java Sometimes when you push the build button to build all op modes, the RC returns an error message that the build failed. If you press the build button a second time, the build typically suceeds. Version 3.2 (built on 17.08.02) This version of the software introduces the "OnBot Java" Development Tool. Similar to the FTC Blocks Development Tool, the FTC OnBot Java Development Tool allows a user to create, edit and build op modes dynamically using only a Javascript-enabled web browser. The OnBot Java Development Tool is an integrated development environment (IDE) that is served up by the Robot Controller. Op modes are created and edited using a Javascript-enabled browser (Google Chromse is recommended). Op modes are saved on the Robot Controller Android device directly. The OnBot Java Development Tool provides a Java programming environment that does NOT need Android Studio. Changes with version 3.2 include: Enhanced web-based development tools Introduction of OnBot Java Development Tool. Web-based programming and management features are "always on" (user no longer needs to put Robot Controller into programming mode). Web-based management interface (where user can change Robot Controller name and also easily download Robot Controller log file). OnBot Java, Blocks and Management features available from web based interface. Blocks Programming Development Tool: Changed "LynxI2cColorRangeSensor" block to "REV Color/range sensor" block. Fixed tooltip for ColorSensor.isLightOn block. Added blocks for ColorSensor.getNormalizedColors and LynxI2cColorRangeSensor.getNormalizedColors. Added example op modes for digital touch sensor and REV Robotics Color Distance sensor. User selectable color themes. Includes many minor enhancements and fixes (too numerous to list). Known issues: Auto complete function is incomplete and does not support the following (for now): Access via this keyword Access via super keyword Members of the super cloass, not overridden by the class Any methods provided in the current class Inner classes Can't handle casted objects Any objects coming from an parenthetically enclosed expression Version 3.10 (built on 17.05.09) This version of the software provides support for the REV Robotics Expansion Hub. This version also includes improvements in the USB communication layer in an effort to enhance system resiliency. If you were using a 2.x version of the software previously, updating to version 3.1 requires that you also update your Driver Station software in addition to updating the Robot Controller software. Also note that in version 3.10 software, the setMaxSpeed and getMaxSpeed methods are no longer available (not deprecated, they have been removed from the SDK). Also note that the the new 3.x software incorporates motor profiles that a user can select as he/she configures the robot. Changes include: Blocks changes Added VuforiaTrackableDefaultListener.getPose and Vuforia.trackPose blocks. Added optimized blocks support for Vuforia extended tracking. Added atan2 block to the math category. Added useCompetitionFieldTargetLocations parameter to Vuforia.initialize block. If set to false, the target locations are placed at (0,0,0) with target orientation as specified in https://github.com/gearsincorg/FTCVuforiaDemo/blob/master/Robot_Navigation.java tutorial op mode. Incorporates additional improvements to USB comm layer to improve system resiliency (to recover from a greater number of communication disruptions). Additional Notes Regarding Version 3.00 (built on 17.04.13) In addition to the release changes listed below (see section labeled "Version 3.00 (built on 17.04.013)"), version 3.00 has the following important changes: Version 3.00 software uses a new version of the FTC Robocol (robot protocol). If you upgrade to v3.0 on the Robot Controller and/or Android Studio side, you must also upgrade the Driver Station software to match the new Robocol. Version 3.00 software removes the setMaxSpeed and getMaxSpeed methods from the DcMotor class. If you have an op mode that formerly used these methods, you will need to remove the references/calls to these methods. Instead, v3.0 provides the max speed information through the use of motor profiles that are selected by the user during robot configuration. Version 3.00 software currently does not have a mechanism to disable extra i2c sensors. We hope to re-introduce this function with a release in the near future. Version 3.00 (built on 17.04.13) *** Use this version of the software at YOUR OWN RISK!!! *** This software is being released as an "alpha" version. Use this version at your own risk! This pre-release software contains SIGNIFICANT changes, including changes to the Wi-Fi Direct pairing mechanism, rewrites of the I2C sensor classes, changes to the USB/FTDI layer, and the introduction of support for the REV Robotics Expansion Hub and the REV Robotics color-range-light sensor. These changes were implemented to improve the reliability and resiliency of the FTC control system. Please note, however, that version 3.00 is considered "alpha" code. This code is being released so that the FIRST community will have an opportunity to test the new REV Expansion Hub electronics module when it becomes available in May. The developers do not recommend using this code for critical applications (i.e., competition use). *** Use this version of the software at YOUR OWN RISK!!! *** Changes include: Major rework of sensor-related infrastructure. Includes rewriting sensor classes to implement synchronous I2C communication. Fix to reset Autonomous timer back to 30 seconds. Implementation of specific motor profiles for approved 12V motors (includes Tetrix, AndyMark, Matrix and REV models). Modest improvements to enhance Wi-Fi P2P pairing. Fixes telemetry log addition race. Publishes all the sources (not just a select few). Includes Block programming improvements Addition of optimized Vuforia blocks. Auto scrollbar to projects and sounds pages. Fixed blocks paste bug. Blocks execute after while-opModeIsActive loop (to allow for cleanup before exiting op mode). Added gyro integratedZValue block. Fixes bug with projects page for Firefox browser. Added IsSpeaking block to AndroidTextToSpeech. Implements support for the REV Robotics Expansion Hub Implements support for integral REV IMU (physically installed on I2C bus 0, uses same Bosch BNO055 9 axis absolute orientation sensor as Adafruit 9DOF abs orientation sensor). - Implements support for REV color/range/light sensor. Provides support to update Expansion Hub firmware through FTC SDK. Detects REV firmware version and records in log file. Includes support for REV Control Hub (note that the REV Control Hub is not yet approved for FTC use). Implements FTC Blocks programming support for REV Expansion Hub and sensor hardware. Detects and alerts when I2C device disconnect. Version 2.62 (built on 17.01.07) Added null pointer check before calling modeToByte() in finishModeSwitchIfNecessary method for ModernRoboticsUsbDcMotorController class. Changes to enhance Modern Robotics USB protocol robustness. Version 2.61 (released on 16.12.19) Blocks Programming mode changes: Fix to correct issue when an exception was thrown because an OpticalDistanceSensor object appears twice in the hardware map (the second time as a LightSensor). Version 2.6 (released on 16.12.16) Fixes for Gyro class: Improve (decrease) sensor refresh latency. fix isCalibrating issues. Blocks Programming mode changes: Blocks now ignores a device in the configuration xml if the name is empty. Other devices work in configuration work fine. Version 2.5 (internal release on released on 16.12.13) Blocks Programming mode changes: Added blocks support for AdafruitBNO055IMU. Added Download Op Mode button to FtcBocks.html. Added support for copying blocks in one OpMode and pasting them in an other OpMode. The clipboard content is stored on the phone, so the programming mode server must be running. Modified Utilities section of the toolbox. In Programming Mode, display information about the active connections. Fixed paste location when workspace has been scrolled. Added blocks support for the android Accelerometer. Fixed issue where Blocks Upload Op Mode truncated name at first dot. Added blocks support for Android SoundPool. Added type safety to blocks for Acceleration. Added type safety to blocks for AdafruitBNO055IMU.Parameters. Added type safety to blocks for AnalogInput. Added type safety to blocks for AngularVelocity. Added type safety to blocks for Color. Added type safety to blocks for ColorSensor. Added type safety to blocks for CompassSensor. Added type safety to blocks for CRServo. Added type safety to blocks for DigitalChannel. Added type safety to blocks for ElapsedTime. Added type safety to blocks for Gamepad. Added type safety to blocks for GyroSensor. Added type safety to blocks for IrSeekerSensor. Added type safety to blocks for LED. Added type safety to blocks for LightSensor. Added type safety to blocks for LinearOpMode. Added type safety to blocks for MagneticFlux. Added type safety to blocks for MatrixF. Added type safety to blocks for MrI2cCompassSensor. Added type safety to blocks for MrI2cRangeSensor. Added type safety to blocks for OpticalDistanceSensor. Added type safety to blocks for Orientation. Added type safety to blocks for Position. Added type safety to blocks for Quaternion. Added type safety to blocks for Servo. Added type safety to blocks for ServoController. Added type safety to blocks for Telemetry. Added type safety to blocks for Temperature. Added type safety to blocks for TouchSensor. Added type safety to blocks for UltrasonicSensor. Added type safety to blocks for VectorF. Added type safety to blocks for Velocity. Added type safety to blocks for VoltageSensor. Added type safety to blocks for VuforiaLocalizer.Parameters. Added type safety to blocks for VuforiaTrackable. Added type safety to blocks for VuforiaTrackables. Added type safety to blocks for enums in AdafruitBNO055IMU.Parameters. Added type safety to blocks for AndroidAccelerometer, AndroidGyroscope, AndroidOrientation, and AndroidTextToSpeech. Version 2.4 (released on 16.11.13) Fix to avoid crashing for nonexistent resources. Blocks Programming mode changes: Added blocks to support OpenGLMatrix, MatrixF, and VectorF. Added blocks to support AngleUnit, AxesOrder, AxesReference, CameraDirection, CameraMonitorFeedback, DistanceUnit, and TempUnit. Added blocks to support Acceleration. Added blocks to support LinearOpMode.getRuntime. Added blocks to support MagneticFlux and Position. Fixed typos. Made blocks for ElapsedTime more consistent with other objects. Added blocks to support Quaternion, Velocity, Orientation, AngularVelocity. Added blocks to support VuforiaTrackables, VuforiaTrackable, VuforiaLocalizer, VuforiaTrackableDefaultListener. Fixed a few blocks. Added type checking to new blocks. Updated to latest blockly. Added default variable blocks to navigation and matrix blocks. Fixed toolbox entry for openGLMatrix_rotation_withAxesArgs. When user downloads Blocks-generated op mode, only the .blk file is downloaded. When user uploads Blocks-generated op mode (.blk file), Javascript code is auto generated. Added DbgLog support. Added logging when a blocks file is read/written. Fixed bug to properly render blocks even if missing devices from configuration file. Added support for additional characters (not just alphanumeric) for the block file names (for download and upload). Added support for OpMode flavor (“Autonomous” or “TeleOp”) and group. Changes to Samples to prevent tutorial issues. Incorporated suggested changes from public pull 216 (“Replace .. paths”). Remove Servo Glitches when robot stopped. if user hits “Cancels” when editing a configuration file, clears the unsaved changes and reverts to original unmodified configuration. Added log info to help diagnose why the Robot Controller app was terminated (for example, by watch dog function). Added ability to transfer log from the controller. Fixed inconsistency for AngularVelocity Limit unbounded growth of data for telemetry. If user does not call telemetry.update() for LinearOpMode in a timely manner, data added for telemetry might get lost if size limit is exceeded. Version 2.35 (released on 16.10.06) Blockly programming mode - Removed unnecesary idle() call from blocks for new project. Version 2.30 (released on 16.10.05) Blockly programming mode: Mechanism added to save Blockly op modes from Programming Mode Server onto local device To avoid clutter, blocks are displayed in categorized folders Added support for DigitalChannel Added support for ModernRoboticsI2cCompassSensor Added support for ModernRoboticsI2cRangeSensor Added support for VoltageSensor Added support for AnalogInput Added support for AnalogOutput Fix for CompassSensor setMode block Vuforia Fix deadlock / make camera data available while Vuforia is running. Update to Vuforia 6.0.117 (recommended by Vuforia and Google to close security loophole). Fix for autonomous 30 second timer bug (where timer was in effect, even though it appeared to have timed out). opModeIsActive changes to allow cleanup after op mode is stopped (with enforced 2 second safety timeout). Fix to avoid reading i2c twice. Updated sample Op Modes. Improved logging and fixed intermittent freezing. Added digital I/O sample. Cleaned up device names in sample op modes to be consistent with Pushbot guide. Fix to allow use of IrSeekerSensorV3. Version 2.20 (released on 16.09.08) Support for Modern Robotics Compass Sensor. Support for Modern Robotics Range Sensor. Revise device names for Pushbot templates to match the names used in Pushbot guide. Fixed bug so that IrSeekerSensorV3 device is accessible as IrSeekerSensor in hardwareMap. Modified computer vision code to require an individual Vuforia license (per legal requirement from PTC). Minor fixes. Blockly enhancements: Support for Voltage Sensor. Support for Analog Input. Support for Analog Output. Support for Light Sensor. Support for Servo Controller. Version 2.10 (released on 16.09.03) Support for Adafruit IMU. Improvements to ModernRoboticsI2cGyro class Block on reset of z axis. isCalibrating() returns true while gyro is calibration. Updated sample gyro program. Blockly enhancements support for android.graphics.Color. added support for ElapsedTime. improved look and legibility of blocks. support for compass sensor. support for ultrasonic sensor. support for IrSeeker. support for LED. support for color sensor. support for CRServo prompt user to configure robot before using programming mode. Provides ability to disable audio cues. various bug fixes and improvements. Version 2.00 (released on 16.08.19) This is the new release for the upcoming 2016-2017 FIRST Tech Challenge Season. Channel change is enabled in the FTC Robot Controller app for Moto G 2nd and 3rd Gen phones. Users can now use annotations to register/disable their Op Modes. Changes in the Android SDK, JDK and build tool requirements (minsdk=19, java 1.7, build tools 23.0.3). Standardized units in analog input. Cleaned up code for existing analog sensor classes. setChannelMode and getChannelMode were REMOVED from the DcMotorController class. This is important - we no longer set the motor modes through the motor controller. setMode and getMode were added to the DcMotor class. ContinuousRotationServo class has been added to the FTC SDK. Range.clip() method has been overloaded so it can support this operation for int, short and byte integers. Some changes have been made (new methods added) on how a user can access items from the hardware map. Users can now set the zero power behavior for a DC motor so that the motor will brake or float when power is zero. Prototype Blockly Programming Mode has been added to FTC Robot Controller. Users can place the Robot Controller into this mode, and then use a device (such as a laptop) that has a Javascript enabled browser to write Blockly-based Op Modes directly onto the Robot Controller. Users can now configure the robot remotely through the FTC Driver Station app. Android Studio project supports Android Studio 2.1.x and compile SDK Version 23 (Marshmallow). Vuforia Computer Vision SDK integrated into FTC SDK. Users can use sample vision targets to get localization information on a standard FTC field. Project structure has been reorganized so that there is now a TeamCode package that users can use to place their local/custom Op Modes into this package. Inspection function has been integrated into the FTC Robot Controller and Driver Station Apps (Thanks Team HazMat… 9277 & 10650!). Audio cues have been incorporated into FTC SDK. Swap mechanism added to FTC Robot Controller configuration activity. For example, if you have two motor controllers on a robot, and you misidentified them in your configuration file, you can use the Swap button to swap the devices within the configuration file (so you do not have to manually re-enter in the configuration info for the two devices). Fix mechanism added to all user to replace an electronic module easily. For example, suppose a servo controller dies on your robot. You replace the broken module with a new module, which has a different serial number from the original servo controller. You can use the Fix button to automatically reconfigure your configuration file to use the serial number of the new module. Improvements made to fix resiliency and responsiveness of the system. For LinearOpMode the user now must for a telemetry.update() to update the telemetry data on the driver station. This update() mechanism ensures that the driver station gets the updated data properly and at the same time. The Auto Configure function of the Robot Controller is now template based. If there is a commonly used robot configuration, a template can be created so that the Auto Configure mechanism can be used to quickly configure a robot of this type. The logic to detect a runaway op mode (both in the LinearOpMode and OpMode types) and to abort the run, then auto recover has been improved/implemented. Fix has been incorporated so that Logitech F310 gamepad mappings will be correct for Marshmallow users. Release 16.07.08 For the ftc_app project, the gradle files have been modified to support Android Studio 2.1.x. Release 16.03.30 For the MIT App Inventor, the design blocks have new icons that better represent the function of each design component. Some changes were made to the shutdown logic to ensure the robust shutdown of some of our USB services. A change was made to LinearOpMode so as to allow a given instance to be executed more than once, which is required for the App Inventor. Javadoc improved/updated. Release 16.03.09 Changes made to make the FTC SDK synchronous (significant change!) waitOneFullHardwareCycle() and waitForNextHardwareCycle() are no longer needed and have been deprecated. runOpMode() (for a LinearOpMode) is now decoupled from the system's hardware read/write thread. loop() (for an OpMode) is now decoupled from the system's hardware read/write thread. Methods are synchronous. For example, if you call setMode(DcMotorController.RunMode.RESET_ENCODERS) for a motor, the encoder is guaranteed to be reset when the method call is complete. For legacy module (NXT compatible), user no longer has to toggle between read and write modes when reading from or writing to a legacy device. Changes made to enhance reliability/robustness during ESD event. Changes made to make code thread safe. Debug keystore added so that user-generated robot controller APKs will all use the same signed key (to avoid conflicts if a team has multiple developer laptops for example). Firmware version information for Modern Robotics modules are now logged. Changes made to improve USB comm reliability and robustness. Added support for voltage indicator for legacy (NXT-compatible) motor controllers. Changes made to provide auto stop capabilities for op modes. A LinearOpMode class will stop when the statements in runOpMode() are complete. User does not have to push the stop button on the driver station. If an op mode is stopped by the driver station, but there is a run away/uninterruptible thread persisting, the app will log an error message then force itself to crash to stop the runaway thread. Driver Station UI modified to display lowest measured voltage below current voltage (12V battery). Driver Station UI modified to have color background for current voltage (green=good, yellow=caution, red=danger, extremely low voltage). javadoc improved (edits and additional classes). Added app build time to About activity for driver station and robot controller apps. Display local IP addresses on Driver Station About activity. Added I2cDeviceSynchImpl. Added I2cDeviceSync interface. Added seconds() and milliseconds() to ElapsedTime for clarity. Added getCallbackCount() to I2cDevice. Added missing clearI2cPortActionFlag. Added code to create log messages while waiting for LinearOpMode shutdown. Fix so Wifi Direct Config activity will no longer launch multiple times. Added the ability to specify an alternate i2c address in software for the Modern Robotics gyro. Release 16.02.09 Improved battery checker feature so that voltage values get refreshed regularly (every 250 msec) on Driver Station (DS) user interface. Improved software so that Robot Controller (RC) is much more resilient and “self-healing” to USB disconnects: If user attempts to start/restart RC with one or more module missing, it will display a warning but still start up. When running an op mode, if one or more modules gets disconnected, the RC & DS will display warnings,and robot will keep on working in spite of the missing module(s). If a disconnected module gets physically reconnected the RC will auto detect the module and the user will regain control of the recently connected module. Warning messages are more helpful (identifies the type of module that’s missing plus its USB serial number). Code changes to fix the null gamepad reference when users try to reference the gamepads in the init() portion of their op mode. NXT light sensor output is now properly scaled. Note that teams might have to readjust their light threshold values in their op modes. On DS user interface, gamepad icon for a driver will disappear if the matching gamepad is disconnected or if that gamepad gets designated as a different driver. Robot Protocol (ROBOCOL) version number info is displayed in About screen on RC and DS apps. Incorporated a display filter on pairing screen to filter out devices that don’t use the “-“ format. This filter can be turned off to show all WiFi Direct devices. Updated text in License file. Fixed formatting error in OpticalDistanceSensor.toString(). Fixed issue on with a blank (“”) device name that would disrupt WiFi Direct Pairing. Made a change so that the WiFi info and battery info can be displayed more quickly on the DS upon connecting to RC. Improved javadoc generation. Modified code to make it easier to support language localization in the future. Release 16.01.04 Updated compileSdkVersion for apps Prevent Wifi from entering power saving mode removed unused import from driver station Corrrected "Dead zone" joystick code. LED.getDeviceName and .getConnectionInfo() return null apps check for ROBOCOL_VERSION mismatch Fix for Telemetry also has off-by-one errors in its data string sizing / short size limitations error User telemetry output is sorted. added formatting variants to DbgLog and RobotLog APIs code modified to allow for a long list of op mode names. changes to improve thread safety of RobocolDatagramSocket Fix for "missing hardware leaves robot controller disconnected from driver station" error fix for "fast tapping of Init/Start causes problems" (toast is now only instantiated on UI thread). added some log statements for thread life cycle. moved gamepad reset logic inside of initActiveOpMode() for robustness changes made to mitigate risk of race conditions on public methods. changes to try and flag when WiFi Direct name contains non-printable characters. fix to correct race condition between .run() and .close() in ReadWriteRunnableStandard. updated FTDI driver made ReadWriteRunnableStanard interface public. fixed off-by-one errors in Command constructor moved specific hardware implmentations into their own package. moved specific gamepad implemnatations to the hardware library. changed LICENSE file to new BSD version. fixed race condition when shutting down Modern Robotics USB devices. methods in the ColorSensor classes have been synchronized. corrected isBusy() status to reflect end of motion. corrected "back" button keycode. the notSupported() method of the GyroSensor class was changed to protected (it should not be public). Release 15.11.04.001 Added Support for Modern Robotics Gyro. The GyroSensor class now supports the MR Gyro Sensor. Users can access heading data (about Z axis) Users can also access raw gyro data (X, Y, & Z axes). Example MRGyroTest.java op mode included. Improved error messages More descriptive error messages for exceptions in user code. Updated DcMotor API Enable read mode on new address in setI2cAddress Fix so that driver station app resets the gamepads when switching op modes. USB-related code changes to make USB comm more responsive and to display more explicit error messages. Fix so that USB will recover properly if the USB bus returns garbage data. Fix USB initializtion race condition. Better error reporting during FTDI open. More explicit messages during USB failures. Fixed bug so that USB device is closed if event loop teardown method was not called. Fixed timer UI issue Fixed duplicate name UI bug (Legacy Module configuration). Fixed race condition in EventLoopManager. Fix to keep references stable when updating gamepad. For legacy Matrix motor/servo controllers removed necessity of appending "Motor" and "Servo" to controller names. Updated HT color sensor driver to use constants from ModernRoboticsUsbLegacyModule class. Updated MR color sensor driver to use constants from ModernRoboticsUsbDeviceInterfaceModule class. Correctly handle I2C Address change in all color sensors Updated/cleaned up op modes. Updated comments in LinearI2cAddressChange.java example op mode. Replaced the calls to "setChannelMode" with "setMode" (to match the new of the DcMotor method). Removed K9AutoTime.java op mode. Added MRGyroTest.java op mode (demonstrates how to use MR Gyro Sensor). Added MRRGBExample.java op mode (demonstrates how to use MR Color Sensor). Added HTRGBExample.java op mode (demonstrates how to use HT legacy color sensor). Added MatrixControllerDemo.java (demonstrates how to use legacy Matrix controller). Updated javadoc documentation. Updated release .apk files for Robot Controller and Driver Station apps. Release 15.10.06.002 Added support for Legacy Matrix 9.6V motor/servo controller. Cleaned up build.gradle file. Minor UI and bug fixes for driver station and robot controller apps. Throws error if Ultrasonic sensor (NXT) is not configured for legacy module port 4 or 5. Release 15.08.03.001 New user interfaces for FTC Driver Station and FTC Robot Controller apps. An init() method is added to the OpMode class. For this release, init() is triggered right before the start() method. Eventually, the init() method will be triggered when the user presses an "INIT" button on driver station. The init() and loop() methods are now required (i.e., need to be overridden in the user's op mode). The start() and stop() methods are optional. A new LinearOpMode class is introduced. Teams can use the LinearOpMode mode to create a linear (not event driven) program model. Teams can use blocking statements like Thread.sleep() within a linear op mode. The API for the Legacy Module and Core Device Interface Module have been updated. Support for encoders with the Legacy Module is now working. The hardware loop has been updated for better performance.
molyswu / Hand Detectionusing Neural Networks (SSD) on Tensorflow. This repo documents steps and scripts used to train a hand detector using Tensorflow (Object Detection API). As with any DNN based task, the most expensive (and riskiest) part of the process has to do with finding or creating the right (annotated) dataset. I was interested mainly in detecting hands on a table (egocentric view point). I experimented first with the [Oxford Hands Dataset](http://www.robots.ox.ac.uk/~vgg/data/hands/) (the results were not good). I then tried the [Egohands Dataset](http://vision.soic.indiana.edu/projects/egohands/) which was a much better fit to my requirements. The goal of this repo/post is to demonstrate how neural networks can be applied to the (hard) problem of tracking hands (egocentric and other views). Better still, provide code that can be adapted to other uses cases. If you use this tutorial or models in your research or project, please cite [this](#citing-this-tutorial). Here is the detector in action. <img src="images/hand1.gif" width="33.3%"><img src="images/hand2.gif" width="33.3%"><img src="images/hand3.gif" width="33.3%"> Realtime detection on video stream from a webcam . <img src="images/chess1.gif" width="33.3%"><img src="images/chess2.gif" width="33.3%"><img src="images/chess3.gif" width="33.3%"> Detection on a Youtube video. Both examples above were run on a macbook pro **CPU** (i7, 2.5GHz, 16GB). Some fps numbers are: | FPS | Image Size | Device| Comments| | ------------- | ------------- | ------------- | ------------- | | 21 | 320 * 240 | Macbook pro (i7, 2.5GHz, 16GB) | Run without visualizing results| | 16 | 320 * 240 | Macbook pro (i7, 2.5GHz, 16GB) | Run while visualizing results (image above) | | 11 | 640 * 480 | Macbook pro (i7, 2.5GHz, 16GB) | Run while visualizing results (image above) | > Note: The code in this repo is written and tested with Tensorflow `1.4.0-rc0`. Using a different version may result in [some errors](https://github.com/tensorflow/models/issues/1581). You may need to [generate your own frozen model](https://pythonprogramming.net/testing-custom-object-detector-tensorflow-object-detection-api-tutorial/?completed=/training-custom-objects-tensorflow-object-detection-api-tutorial/) graph using the [model checkpoints](model-checkpoint) in the repo to fit your TF version. **Content of this document** - Motivation - Why Track/Detect hands with Neural Networks - Data preparation and network training in Tensorflow (Dataset, Import, Training) - Training the hand detection Model - Using the Detector to Detect/Track hands - Thoughts on Optimizations. > P.S if you are using or have used the models provided here, feel free to reach out on twitter ([@vykthur](https://twitter.com/vykthur)) and share your work! ## Motivation - Why Track/Detect hands with Neural Networks? There are several existing approaches to tracking hands in the computer vision domain. Incidentally, many of these approaches are rule based (e.g extracting background based on texture and boundary features, distinguishing between hands and background using color histograms and HOG classifiers,) making them not very robust. For example, these algorithms might get confused if the background is unusual or in situations where sharp changes in lighting conditions cause sharp changes in skin color or the tracked object becomes occluded.(see [here for a review](https://www.cse.unr.edu/~bebis/handposerev.pdf) paper on hand pose estimation from the HCI perspective) With sufficiently large datasets, neural networks provide opportunity to train models that perform well and address challenges of existing object tracking/detection algorithms - varied/poor lighting, noisy environments, diverse viewpoints and even occlusion. The main drawbacks to usage for real-time tracking/detection is that they can be complex, are relatively slow compared to tracking-only algorithms and it can be quite expensive to assemble a good dataset. But things are changing with advances in fast neural networks. Furthermore, this entire area of work has been made more approachable by deep learning frameworks (such as the tensorflow object detection api) that simplify the process of training a model for custom object detection. More importantly, the advent of fast neural network models like ssd, faster r-cnn, rfcn (see [here](https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/detection_model_zoo.md#coco-trained-models-coco-models) ) etc make neural networks an attractive candidate for real-time detection (and tracking) applications. Hopefully, this repo demonstrates this. > If you are not interested in the process of training the detector, you can skip straight to applying the [pretrained model I provide in detecting hands](#detecting-hands). Training a model is a multi-stage process (assembling dataset, cleaning, splitting into training/test partitions and generating an inference graph). While I lightly touch on the details of these parts, there are a few other tutorials cover training a custom object detector using the tensorflow object detection api in more detail[ see [here](https://pythonprogramming.net/training-custom-objects-tensorflow-object-detection-api-tutorial/) and [here](https://towardsdatascience.com/how-to-train-your-own-object-detector-with-tensorflows-object-detector-api-bec72ecfe1d9) ]. I recommend you walk through those if interested in training a custom object detector from scratch. ## Data preparation and network training in Tensorflow (Dataset, Import, Training) **The Egohands Dataset** The hand detector model is built using data from the [Egohands Dataset](http://vision.soic.indiana.edu/projects/egohands/) dataset. This dataset works well for several reasons. It contains high quality, pixel level annotations (>15000 ground truth labels) where hands are located across 4800 images. All images are captured from an egocentric view (Google glass) across 48 different environments (indoor, outdoor) and activities (playing cards, chess, jenga, solving puzzles etc). <img src="images/egohandstrain.jpg" width="100%"> If you will be using the Egohands dataset, you can cite them as follows: > Bambach, Sven, et al. "Lending a hand: Detecting hands and recognizing activities in complex egocentric interactions." Proceedings of the IEEE International Conference on Computer Vision. 2015. The Egohands dataset (zip file with labelled data) contains 48 folders of locations where video data was collected (100 images per folder). ``` -- LOCATION_X -- frame_1.jpg -- frame_2.jpg ... -- frame_100.jpg -- polygons.mat // contains annotations for all 100 images in current folder -- LOCATION_Y -- frame_1.jpg -- frame_2.jpg ... -- frame_100.jpg -- polygons.mat // contains annotations for all 100 images in current folder ``` **Converting data to Tensorflow Format** Some initial work needs to be done to the Egohands dataset to transform it into the format (`tfrecord`) which Tensorflow needs to train a model. This repo contains `egohands_dataset_clean.py` a script that will help you generate these csv files. - Downloads the egohands datasets - Renames all files to include their directory names to ensure each filename is unique - Splits the dataset into train (80%), test (10%) and eval (10%) folders. - Reads in `polygons.mat` for each folder, generates bounding boxes and visualizes them to ensure correctness (see image above). - Once the script is done running, you should have an images folder containing three folders - train, test and eval. Each of these folders should also contain a csv label document each - `train_labels.csv`, `test_labels.csv` that can be used to generate `tfrecords` Note: While the egohands dataset provides four separate labels for hands (own left, own right, other left, and other right), for my purpose, I am only interested in the general `hand` class and label all training data as `hand`. You can modify the data prep script to generate `tfrecords` that support 4 labels. Next: convert your dataset + csv files to tfrecords. A helpful guide on this can be found [here](https://pythonprogramming.net/creating-tfrecord-files-tensorflow-object-detection-api-tutorial/).For each folder, you should be able to generate `train.record`, `test.record` required in the training process. ## Training the hand detection Model Now that the dataset has been assembled (and your tfrecords), the next task is to train a model based on this. With neural networks, it is possible to use a process called [transfer learning](https://www.tensorflow.org/tutorials/image_retraining) to shorten the amount of time needed to train the entire model. This means we can take an existing model (that has been trained well on a related domain (here image classification) and retrain its final layer(s) to detect hands for us. Sweet!. Given that neural networks sometimes have thousands or millions of parameters that can take weeks or months to train, transfer learning helps shorten training time to possibly hours. Tensorflow does offer a few models (in the tensorflow [model zoo](https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/detection_model_zoo.md#coco-trained-models-coco-models)) and I chose to use the `ssd_mobilenet_v1_coco` model as my start point given it is currently (one of) the fastest models (read the SSD research [paper here](https://arxiv.org/pdf/1512.02325.pdf)). The training process can be done locally on your CPU machine which may take a while or better on a (cloud) GPU machine (which is what I did). For reference, training on my macbook pro (tensorflow compiled from source to take advantage of the mac's cpu architecture) the maximum speed I got was 5 seconds per step as opposed to the ~0.5 seconds per step I got with a GPU. For reference it would take about 12 days to run 200k steps on my mac (i7, 2.5GHz, 16GB) compared to ~5hrs on a GPU. > **Training on your own images**: Please use the [guide provided by Harrison from pythonprogramming](https://pythonprogramming.net/training-custom-objects-tensorflow-object-detection-api-tutorial/) on how to generate tfrecords given your label csv files and your images. The guide also covers how to start the training process if training locally. [see [here] (https://pythonprogramming.net/training-custom-objects-tensorflow-object-detection-api-tutorial/)]. If training in the cloud using a service like GCP, see the [guide here](https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/running_on_cloud.md). As the training process progresses, the expectation is that total loss (errors) gets reduced to its possible minimum (about a value of 1 or thereabout). By observing the tensorboard graphs for total loss(see image below), it should be possible to get an idea of when the training process is complete (total loss does not decrease with further iterations/steps). I ran my training job for 200k steps (took about 5 hours) and stopped at a total Loss (errors) value of 2.575.(In retrospect, I could have stopped the training at about 50k steps and gotten a similar total loss value). With tensorflow, you can also run an evaluation concurrently that assesses your model to see how well it performs on the test data. A commonly used metric for performance is mean average precision (mAP) which is single number used to summarize the area under the precision-recall curve. mAP is a measure of how well the model generates a bounding box that has at least a 50% overlap with the ground truth bounding box in our test dataset. For the hand detector trained here, the mAP value was **0.9686@0.5IOU**. mAP values range from 0-1, the higher the better. <img src="images/accuracy.jpg" width="100%"> Once training is completed, the trained inference graph (`frozen_inference_graph.pb`) is then exported (see the earlier referenced guides for how to do this) and saved in the `hand_inference_graph` folder. Now its time to do some interesting detection. ## Using the Detector to Detect/Track hands If you have not done this yet, please following the guide on installing [Tensorflow and the Tensorflow object detection api](https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/installation.md). This will walk you through setting up the tensorflow framework, cloning the tensorflow github repo and a guide on - Load the `frozen_inference_graph.pb` trained on the hands dataset as well as the corresponding label map. In this repo, this is done in the `utils/detector_utils.py` script by the `load_inference_graph` method. ```python detection_graph = tf.Graph() with detection_graph.as_default(): od_graph_def = tf.GraphDef() with tf.gfile.GFile(PATH_TO_CKPT, 'rb') as fid: serialized_graph = fid.read() od_graph_def.ParseFromString(serialized_graph) tf.import_graph_def(od_graph_def, name='') sess = tf.Session(graph=detection_graph) print("> ====== Hand Inference graph loaded.") ``` - Detect hands. In this repo, this is done in the `utils/detector_utils.py` script by the `detect_objects` method. ```python (boxes, scores, classes, num) = sess.run( [detection_boxes, detection_scores, detection_classes, num_detections], feed_dict={image_tensor: image_np_expanded}) ``` - Visualize detected bounding detection_boxes. In this repo, this is done in the `utils/detector_utils.py` script by the `draw_box_on_image` method. This repo contains two scripts that tie all these steps together. - detect_multi_threaded.py : A threaded implementation for reading camera video input detection and detecting. Takes a set of command line flags to set parameters such as `--display` (visualize detections), image parameters `--width` and `--height`, videe `--source` (0 for camera) etc. - detect_single_threaded.py : Same as above, but single threaded. This script works for video files by setting the video source parameter videe `--source` (path to a video file). ```cmd # load and run detection on video at path "videos/chess.mov" python detect_single_threaded.py --source videos/chess.mov ``` > Update: If you do have errors loading the frozen inference graph in this repo, feel free to generate a new graph that fits your TF version from the model-checkpoint in this repo. Use the [export_inference_graph.py](https://github.com/tensorflow/models/blob/master/research/object_detection/export_inference_graph.py) script provided in the tensorflow object detection api repo. More guidance on this [here](https://pythonprogramming.net/testing-custom-object-detector-tensorflow-object-detection-api-tutorial/?completed=/training-custom-objects-tensorflow-object-detection-api-tutorial/). ## Thoughts on Optimization. A few things that led to noticeable performance increases. - Threading: Turns out that reading images from a webcam is a heavy I/O event and if run on the main application thread can slow down the program. I implemented some good ideas from [Adrian Rosebuck](https://www.pyimagesearch.com/2017/02/06/faster-video-file-fps-with-cv2-videocapture-and-opencv/) on parrallelizing image capture across multiple worker threads. This mostly led to an FPS increase of about 5 points. - For those new to Opencv, images from the `cv2.read()` method return images in [BGR format](https://www.learnopencv.com/why-does-opencv-use-bgr-color-format/). Ensure you convert to RGB before detection (accuracy will be much reduced if you dont). ```python cv2.cvtColor(image_np, cv2.COLOR_BGR2RGB) ``` - Keeping your input image small will increase fps without any significant accuracy drop.(I used about 320 x 240 compared to the 1280 x 720 which my webcam provides). - Model Quantization. Moving from the current 32 bit to 8 bit can achieve up to 4x reduction in memory required to load and store models. One way to further speed up this model is to explore the use of [8-bit fixed point quantization](https://heartbeat.fritz.ai/8-bit-quantization-and-tensorflow-lite-speeding-up-mobile-inference-with-low-precision-a882dfcafbbd). Performance can also be increased by a clever combination of tracking algorithms with the already decent detection and this is something I am still experimenting with. Have ideas for optimizing better, please share! <img src="images/general.jpg" width="100%"> Note: The detector does reflect some limitations associated with the training set. This includes non-egocentric viewpoints, very noisy backgrounds (e.g in a sea of hands) and sometimes skin tone. There is opportunity to improve these with additional data. ## Integrating Multiple DNNs. One way to make things more interesting is to integrate our new knowledge of where "hands" are with other detectors trained to recognize other objects. Unfortunately, while our hand detector can in fact detect hands, it cannot detect other objects (a factor or how it is trained). To create a detector that classifies multiple different objects would mean a long involved process of assembling datasets for each class and a lengthy training process. > Given the above, a potential strategy is to explore structures that allow us **efficiently** interleave output form multiple pretrained models for various object classes and have them detect multiple objects on a single image. An example of this is with my primary use case where I am interested in understanding the position of objects on a table with respect to hands on same table. I am currently doing some work on a threaded application that loads multiple detectors and outputs bounding boxes on a single image. More on this soon.
phildougherty / Qwen2.5 VL Inference OpenaiInference service for Qwen2.5-VL-7b model
vtuber-plan / LangportLangport is a language model inference service
Aryia-Behroziuan / NeuronsAn ANN is a model based on a collection of connected units or nodes called "artificial neurons", which loosely model the neurons in a biological brain. Each connection, like the synapses in a biological brain, can transmit information, a "signal", from one artificial neuron to another. An artificial neuron that receives a signal can process it and then signal additional artificial neurons connected to it. In common ANN implementations, the signal at a connection between artificial neurons is a real number, and the output of each artificial neuron is computed by some non-linear function of the sum of its inputs. The connections between artificial neurons are called "edges". Artificial neurons and edges typically have a weight that adjusts as learning proceeds. The weight increases or decreases the strength of the signal at a connection. Artificial neurons may have a threshold such that the signal is only sent if the aggregate signal crosses that threshold. Typically, artificial neurons are aggregated into layers. Different layers may perform different kinds of transformations on their inputs. Signals travel from the first layer (the input layer) to the last layer (the output layer), possibly after traversing the layers multiple times. The original goal of the ANN approach was to solve problems in the same way that a human brain would. However, over time, attention moved to performing specific tasks, leading to deviations from biology. Artificial neural networks have been used on a variety of tasks, including computer vision, speech recognition, machine translation, social network filtering, playing board and video games and medical diagnosis. Deep learning consists of multiple hidden layers in an artificial neural network. This approach tries to model the way the human brain processes light and sound into vision and hearing. Some successful applications of deep learning are computer vision and speech recognition.[68] Decision trees Main article: Decision tree learning Decision tree learning uses a decision tree as a predictive model to go from observations about an item (represented in the branches) to conclusions about the item's target value (represented in the leaves). It is one of the predictive modeling approaches used in statistics, data mining, and machine learning. Tree models where the target variable can take a discrete set of values are called classification trees; in these tree structures, leaves represent class labels and branches represent conjunctions of features that lead to those class labels. Decision trees where the target variable can take continuous values (typically real numbers) are called regression trees. In decision analysis, a decision tree can be used to visually and explicitly represent decisions and decision making. In data mining, a decision tree describes data, but the resulting classification tree can be an input for decision making. Support vector machines Main article: Support vector machines Support vector machines (SVMs), also known as support vector networks, are a set of related supervised learning methods used for classification and regression. Given a set of training examples, each marked as belonging to one of two categories, an SVM training algorithm builds a model that predicts whether a new example falls into one category or the other.[69] An SVM training algorithm is a non-probabilistic, binary, linear classifier, although methods such as Platt scaling exist to use SVM in a probabilistic classification setting. In addition to performing linear classification, SVMs can efficiently perform a non-linear classification using what is called the kernel trick, implicitly mapping their inputs into high-dimensional feature spaces. Illustration of linear regression on a data set. Regression analysis Main article: Regression analysis Regression analysis encompasses a large variety of statistical methods to estimate the relationship between input variables and their associated features. Its most common form is linear regression, where a single line is drawn to best fit the given data according to a mathematical criterion such as ordinary least squares. The latter is often extended by regularization (mathematics) methods to mitigate overfitting and bias, as in ridge regression. When dealing with non-linear problems, go-to models include polynomial regression (for example, used for trendline fitting in Microsoft Excel[70]), logistic regression (often used in statistical classification) or even kernel regression, which introduces non-linearity by taking advantage of the kernel trick to implicitly map input variables to higher-dimensional space. Bayesian networks Main article: Bayesian network A simple Bayesian network. Rain influences whether the sprinkler is activated, and both rain and the sprinkler influence whether the grass is wet. A Bayesian network, belief network, or directed acyclic graphical model is a probabilistic graphical model that represents a set of random variables and their conditional independence with a directed acyclic graph (DAG). For example, a Bayesian network could represent the probabilistic relationships between diseases and symptoms. Given symptoms, the network can be used to compute the probabilities of the presence of various diseases. Efficient algorithms exist that perform inference and learning. Bayesian networks that model sequences of variables, like speech signals or protein sequences, are called dynamic Bayesian networks. Generalizations of Bayesian networks that can represent and solve decision problems under uncertainty are called influence diagrams. Genetic algorithms Main article: Genetic algorithm A genetic algorithm (GA) is a search algorithm and heuristic technique that mimics the process of natural selection, using methods such as mutation and crossover to generate new genotypes in the hope of finding good solutions to a given problem. In machine learning, genetic algorithms were used in the 1980s and 1990s.[71][72] Conversely, machine learning techniques have been used to improve the performance of genetic and evolutionary algorithms.[73] Training models Usually, machine learning models require a lot of data in order for them to perform well. Usually, when training a machine learning model, one needs to collect a large, representative sample of data from a training set. Data from the training set can be as varied as a corpus of text, a collection of images, and data collected from individual users of a service. Overfitting is something to watch out for when training a machine learning model. Federated learning Main article: Federated learning Federated learning is an adapted form of distributed artificial intelligence to training machine learning models that decentralizes the training process, allowing for users' privacy to be maintained by not needing to send their data to a centralized server. This also increases efficiency by decentralizing the training process to many devices. For example, Gboard uses federated machine learning to train search query prediction models on users' mobile phones without having to send individual searches back to Google.[74] Applications There are many applications for machine learning, including: Agriculture Anatomy Adaptive websites Affective computing Banking Bioinformatics Brain–machine interfaces Cheminformatics Citizen science Computer networks Computer vision Credit-card fraud detection Data quality DNA sequence classification Economics Financial market analysis[75] General game playing Handwriting recognition Information retrieval Insurance Internet fraud detection Linguistics Machine learning control Machine perception Machine translation Marketing Medical diagnosis Natural language processing Natural language understanding Online advertising Optimization Recommender systems Robot locomotion Search engines Sentiment analysis Sequence mining Software engineering Speech recognition Structural health monitoring Syntactic pattern recognition Telecommunication Theorem proving Time series forecasting User behavior analytics In 2006, the media-services provider Netflix held the first "Netflix Prize" competition to find a program to better predict user preferences and improve the accuracy of its existing Cinematch movie recommendation algorithm by at least 10%. A joint team made up of researchers from AT&T Labs-Research in collaboration with the teams Big Chaos and Pragmatic Theory built an ensemble model to win the Grand Prize in 2009 for $1 million.[76] Shortly after the prize was awarded, Netflix realized that viewers' ratings were not the best indicators of their viewing patterns ("everything is a recommendation") and they changed their recommendation engine accordingly.[77] In 2010 The Wall Street Journal wrote about the firm Rebellion Research and their use of machine learning to predict the financial crisis.[78] In 2012, co-founder of Sun Microsystems, Vinod Khosla, predicted that 80% of medical doctors' jobs would be lost in the next two decades to automated machine learning medical diagnostic software.[79] In 2014, it was reported that a machine learning algorithm had been applied in the field of art history to study fine art paintings and that it may have revealed previously unrecognized influences among artists.[80] In 2019 Springer Nature published the first research book created using machine learning.[81] Limitations Although machine learning has been transformative in some fields, machine-learning programs often fail to deliver expected results.[82][83][84] Reasons for this are numerous: lack of (suitable) data, lack of access to the data, data bias, privacy problems, badly chosen tasks and algorithms, wrong tools and people, lack of resources, and evaluation problems.[85] In 2018, a self-driving car from Uber failed to detect a pedestrian, who was killed after a collision.[86] Attempts to use machine learning in healthcare with the IBM Watson system failed to deliver even after years of time and billions of dollars invested.[87][88] Bias Main article: Algorithmic bias Machine learning approaches in particular can suffer from different data biases. A machine learning system trained on current customers only may not be able to predict the needs of new customer groups that are not represented in the training data. When trained on man-made data, machine learning is likely to pick up the same constitutional and unconscious biases already present in society.[89] Language models learned from data have been shown to contain human-like biases.[90][91] Machine learning systems used for criminal risk assessment have been found to be biased against black people.[92][93] In 2015, Google photos would often tag black people as gorillas,[94] and in 2018 this still was not well resolved, but Google reportedly was still using the workaround to remove all gorillas from the training data, and thus was not able to recognize real gorillas at all.[95] Similar issues with recognizing non-white people have been found in many other systems.[96] In 2016, Microsoft tested a chatbot that learned from Twitter, and it quickly picked up racist and sexist language.[97] Because of such challenges, the effective use of machine learning may take longer to be adopted in other domains.[98] Concern for fairness in machine learning, that is, reducing bias in machine learning and propelling its use for human good is increasingly expressed by artificial intelligence scientists, including Fei-Fei Li, who reminds engineers that "There’s nothing artificial about AI...It’s inspired by people, it’s created by people, and—most importantly—it impacts people. It is a powerful tool we are only just beginning to understand, and that is a profound responsibility.”[99] Model assessments Classification of machine learning models can be validated by accuracy estimation techniques like the holdout method, which splits the data in a training and test set (conventionally 2/3 training set and 1/3 test set designation) and evaluates the performance of the training model on the test set. In comparison, the K-fold-cross-validation method randomly partitions the data into K subsets and then K experiments are performed each respectively considering 1 subset for evaluation and the remaining K-1 subsets for training the model. In addition to the holdout and cross-validation methods, bootstrap, which samples n instances with replacement from the dataset, can be used to assess model accuracy.[100] In addition to overall accuracy, investigators frequently report sensitivity and specificity meaning True Positive Rate (TPR) and True Negative Rate (TNR) respectively. Similarly, investigators sometimes report the false positive rate (FPR) as well as the false negative rate (FNR). However, these rates are ratios that fail to reveal their numerators and denominators. The total operating characteristic (TOC) is an effective method to express a model's diagnostic ability. TOC shows the numerators and denominators of the previously mentioned rates, thus TOC provides more information than the commonly used receiver operating characteristic (ROC) and ROC's associated area under the curve (AUC).[101] Ethics Machine learning poses a host of ethical questions. Systems which are trained on datasets collected with biases may exhibit these biases upon use (algorithmic bias), thus digitizing cultural prejudices.[102] For example, using job hiring data from a firm with racist hiring policies may lead to a machine learning system duplicating the bias by scoring job applicants against similarity to previous successful applicants.[103][104] Responsible collection of data and documentation of algorithmic rules used by a system thus is a critical part of machine learning. Because human languages contain biases, machines trained on language corpora will necessarily also learn these biases.[105][106] Other forms of ethical challenges, not related to personal biases, are more seen in health care. There are concerns among health care professionals that these systems might not be designed in the public's interest but as income-generating machines. This is especially true in the United States where there is a long-standing ethical dilemma of improving health care, but also increasing profits. For example, the algorithms could be designed to provide patients with unnecessary tests or medication in which the algorithm's proprietary owners hold stakes. There is huge potential for machine learning in health care to provide professionals a great tool to diagnose, medicate, and even plan recovery paths for patients, but this will not happen until the personal biases mentioned previously, and these "greed" biases are addressed.[107] Hardware Since the 2010s, advances in both machine learning algorithms and computer hardware have led to more efficient methods for training deep neural networks (a particular narrow subdomain of machine learning) that contain many layers of non-linear hidden units.[108] By 2019, graphic processing units (GPUs), often with AI-specific enhancements, had displaced CPUs as the dominant method of training large-scale commercial cloud AI.[109] OpenAI estimated the hardware compute used in the largest deep learning projects from AlexNet (2012) to AlphaZero (2017), and found a 300,000-fold increase in the amount of compute required, with a doubling-time trendline of 3.4 months.[110][111] Software Software suites containing a variety of machine learning algorithms include the following: Free and open-source so
bentoml / CLIP API ServiceCLIP as a service - Embed image and sentences, object recognition, visual reasoning, image classification and reverse image search
0xstely / Rust Inference ServiceInference service built in Rust for deploying Large Language Models
aws-solutions-library-samples / Guidance For Machine Learning Inference On AwsThis Guidance demonstrates how to deploy a machine learning inference architecture on Amazon Elastic Kubernetes Service (Amazon EKS). It addresses the basic implementation requirements as well as ways you can pack thousands of unique PyTorch deep learning (DL) models into a scalable architecture and evaluate performance
bentoml / Transformers Nlp ServiceOnline Inference API for NLP Transformer models - summarization, text classification, sentiment analysis and more
opea-project / Enterprise InferenceIntel® AI for Enterprise Inference optimizes AI inference services on Intel hardware using Kubernetes Orchestration. It automates LLM model deployment for faster inference, resource provisioning, and optimal settings to simplify processes and reduce manual work.
aws-samples / Aws Serverless For Machine Learning InferenceThis is a sample solution for bringing your own ML models and inference code, and running them at scale using AWS serverless services.
mudigosa / Image ClassifierImage Classifier Going forward, AI algorithms will be incorporated into more and more everyday applications. For example, you might want to include an image classifier in a smartphone app. To do this, you'd use a deep learning model trained on hundreds of thousands of images as part of the overall application architecture. A large part of software development in the future will be using these types of models as common parts of applications. In this project, you'll train an image classifier to recognize different species of flowers. You can imagine using something like this in a phone app that tells you the name of the flower your camera is looking at. In practice, you'd train this classifier, then export it for use in your application. We'll be using this dataset of 102 flower categories. When you've completed this project, you'll have an application that can be trained on any set of labelled images. Here your network will be learning about flowers and end up as a command line application. But, what you do with your new skills depends on your imagination and effort in building a dataset. This is the final Project of the Udacity AI with Python Nanodegree Prerequisites The Code is written in Python 3.6.5 . If you don't have Python installed you can find it here. If you are using a lower version of Python you can upgrade using the pip package, ensuring you have the latest version of pip. To install pip run in the command Line python -m ensurepip -- default-pip to upgrade it python -m pip install -- upgrade pip setuptools wheel to upgrade Python pip install python -- upgrade Additional Packages that are required are: Numpy, Pandas, MatplotLib, Pytorch, PIL and json. You can donwload them using pip pip install numpy pandas matplotlib pil or conda conda install numpy pandas matplotlib pil In order to intall Pytorch head over to the Pytorch site select your specs and follow the instructions given. Viewing the Jyputer Notebook In order to better view and work on the jupyter Notebook I encourage you to use nbviewer . You can simply copy and paste the link to this website and you will be able to edit it without any problem. Alternatively you can clone the repository using git clone https://github.com/fotisk07/Image-Classifier/ then in the command Line type, after you have downloaded jupyter notebook type jupyter notebook locate the notebook and run it. Command Line Application Train a new network on a data set with train.py Basic Usage : python train.py data_directory Prints out current epoch, training loss, validation loss, and validation accuracy as the netowrk trains Options: Set direcotry to save checkpoints: python train.py data_dor --save_dir save_directory Choose arcitecture (alexnet, densenet121 or vgg16 available): pytnon train.py data_dir --arch "vgg16" Set hyperparameters: python train.py data_dir --learning_rate 0.001 --hidden_layer1 120 --epochs 20 Use GPU for training: python train.py data_dir --gpu gpu Predict flower name from an image with predict.py along with the probability of that name. That is you'll pass in a single image /path/to/image and return the flower name and class probability Basic usage: python predict.py /path/to/image checkpoint Options: Return top K most likely classes: python predict.py input checkpoint ---top_k 3 Use a mapping of categories to real names: python predict.py input checkpoint --category_names cat_To_name.json Use GPU for inference: python predict.py input checkpoint --gpu Json file In order for the network to print out the name of the flower a .json file is required. If you aren't familiar with json you can find information here. By using a .json file the data can be sorted into folders with numbers and those numbers will correspond to specific names specified in the .json file. Data and the json file The data used specifically for this assignemnt are a flower database are not provided in the repository as it's larger than what github allows. Nevertheless, feel free to create your own databases and train the model on them to use with your own projects. The structure of your data should be the following: The data need to comprised of 3 folders, test, train and validate. Generally the proportions should be 70% training 10% validate and 20% test. Inside the train, test and validate folders there should be folders bearing a specific number which corresponds to a specific category, clarified in the json file. For example if we have the image a.jpj and it is a rose it could be in a path like this /test/5/a.jpg and json file would be like this {...5:"rose",...}. Make sure to include a lot of photos of your catagories (more than 10) with different angles and different lighting conditions in order for the network to generalize better. GPU As the network makes use of a sophisticated deep convolutional neural network the training process is impossible to be done by a common laptop. In order to train your models to your local machine you have three options Cuda -- If you have an NVIDIA GPU then you can install CUDA from here. With Cuda you will be able to train your model however the process will still be time consuming Cloud Services -- There are many paid cloud services that let you train your models like AWS or Google Cloud Coogle Colab -- Google Colab gives you free access to a tesla K80 GPU for 12 hours at a time. Once 12 hours have ellapsed you can just reload and continue! The only limitation is that you have to upload the data to Google Drive and if the dataset is massive you may run out of space. However, once a model is trained then a normal CPU can be used for the predict.py file and you will have an answer within some seconds. Hyperparameters As you can see you have a wide selection of hyperparameters available and you can get even more by making small modifications to the code. Thus it may seem overly complicated to choose the right ones especially if the training needs at least 15 minutes to be completed. So here are some hints: By increasing the number of epochs the accuracy of the network on the training set gets better and better however be careful because if you pick a large number of epochs the network won't generalize well, that is to say it will have high accuracy on the training image and low accuracy on the test images. Eg: training for 12 epochs training accuracy: 85% Test accuracy: 82%. Training for 30 epochs training accuracy 95% test accuracy 50%. A big learning rate guarantees that the network will converge fast to a small error but it will constantly overshot A small learning rate guarantees that the network will reach greater accuracies but the learning process will take longer Densenet121 works best for images but the training process takes significantly longer than alexnet or vgg16 *My settings were lr=0.001, dropoup=0.5, epochs= 15 and my test accuracy was 86% with densenet121 as my feature extraction model. Pre-Trained Network The checkpoint.pth file contains the information of a network trained to recognise 102 different species of flowers. I has been trained with specific hyperparameters thus if you don't set them right the network will fail. In order to have a prediction for an image located in the path /path/to/image using my pretrained model you can simply type python predict.py /path/to/image checkpoint.pth Contributing Please read CONTRIBUTING.md for the process for submitting pull requests. Authors Shanmukha Mudigonda - Initial work Udacity - Final Project of the AI with Python Nanodegree
Aryia-Behroziuan / Other SourcesAsada, M.; Hosoda, K.; Kuniyoshi, Y.; Ishiguro, H.; Inui, T.; Yoshikawa, Y.; Ogino, M.; Yoshida, C. (2009). "Cognitive developmental robotics: a survey". IEEE Transactions on Autonomous Mental Development. 1 (1): 12–34. doi:10.1109/tamd.2009.2021702. S2CID 10168773. "ACM Computing Classification System: Artificial intelligence". ACM. 1998. Archived from the original on 12 October 2007. Retrieved 30 August 2007. Goodman, Joanna (2016). Robots in Law: How Artificial Intelligence is Transforming Legal Services (1st ed.). Ark Group. ISBN 978-1-78358-264-8. Archived from the original on 8 November 2016. Retrieved 7 November 2016. Albus, J. S. (2002). "4-D/RCS: A Reference Model Architecture for Intelligent Unmanned Ground Vehicles" (PDF). In Gerhart, G.; Gunderson, R.; Shoemaker, C. (eds.). Proceedings of the SPIE AeroSense Session on Unmanned Ground Vehicle Technology. Unmanned Ground Vehicle Technology IV. 3693. pp. 11–20. Bibcode:2002SPIE.4715..303A. CiteSeerX 10.1.1.15.14. doi:10.1117/12.474462. S2CID 63339739. Archived from the original (PDF) on 25 July 2004. Aleksander, Igor (1995). Artificial Neuroconsciousness: An Update. IWANN. Archived from the original on 2 March 1997. BibTex Archived 2 March 1997 at the Wayback Machine. Bach, Joscha (2008). "Seven Principles of Synthetic Intelligence". In Wang, Pei; Goertzel, Ben; Franklin, Stan (eds.). Artificial General Intelligence, 2008: Proceedings of the First AGI Conference. IOS Press. pp. 63–74. ISBN 978-1-58603-833-5. Archived from the original on 8 July 2016. Retrieved 16 February 2016. "Robots could demand legal rights". BBC News. 21 December 2006. Archived from the original on 15 October 2019. Retrieved 3 February 2011. Brooks, Rodney (1990). "Elephants Don't Play Chess" (PDF). Robotics and Autonomous Systems. 6 (1–2): 3–15. CiteSeerX 10.1.1.588.7539. doi:10.1016/S0921-8890(05)80025-9. Archived (PDF) from the original on 9 August 2007. Brooks, R. A. (1991). "How to build complete creatures rather than isolated cognitive simulators". In VanLehn, K. (ed.). Architectures for Intelligence. Hillsdale, NJ: Lawrence Erlbaum Associates. pp. 225–239. CiteSeerX 10.1.1.52.9510. Buchanan, Bruce G. (2005). "A (Very) Brief History of Artificial Intelligence" (PDF). AI Magazine: 53–60. Archived from the original (PDF) on 26 September 2007. Butler, Samuel (13 June 1863). "Darwin among the Machines". Letters to the Editor. The Press. Christchurch, New Zealand. Archived from the original on 19 September 2008. Retrieved 16 October 2014 – via Victoria University of Wellington. Clark, Jack (8 December 2015). "Why 2015 Was a Breakthrough Year in Artificial Intelligence". Bloomberg News. Archived from the original on 23 November 2016. Retrieved 23 November 2016. After a half-decade of quiet breakthroughs in artificial intelligence, 2015 has been a landmark year. Computers are smarter and learning faster than ever. "AI set to exceed human brain power". CNN. 26 July 2006. Archived from the original on 19 February 2008. Dennett, Daniel (1991). Consciousness Explained. The Penguin Press. ISBN 978-0-7139-9037-9. Domingos, Pedro (2015). The Master Algorithm: How the Quest for the Ultimate Learning Machine Will Remake Our World. Basic Books. ISBN 978-0-465-06192-1. Dowe, D. L.; Hajek, A. R. (1997). "A computational extension to the Turing Test". Proceedings of the 4th Conference of the Australasian Cognitive Science Society. Archived from the original on 28 June 2011. Dreyfus, Hubert (1972). What Computers Can't Do. New York: MIT Press. ISBN 978-0-06-011082-6. Dreyfus, Hubert; Dreyfus, Stuart (1986). Mind over Machine: The Power of Human Intuition and Expertise in the Era of the Computer. Oxford, UK: Blackwell. ISBN 978-0-02-908060-3. Archived from the original on 26 July 2020. Retrieved 22 August 2020. Dreyfus, Hubert (1992). What Computers Still Can't Do. New York: MIT Press. ISBN 978-0-262-54067-4. Dyson, George (1998). Darwin among the Machines. Allan Lane Science. ISBN 978-0-7382-0030-9. Archived from the original on 26 July 2020. Retrieved 22 August 2020. Edelman, Gerald (23 November 2007). "Gerald Edelman – Neural Darwinism and Brain-based Devices". Talking Robots. Archived from the original on 8 October 2009. Edelson, Edward (1991). The Nervous System. New York: Chelsea House. ISBN 978-0-7910-0464-7. Archived from the original on 26 July 2020. Retrieved 18 November 2019. Fearn, Nicholas (2007). The Latest Answers to the Oldest Questions: A Philosophical Adventure with the World's Greatest Thinkers. New York: Grove Press. ISBN 978-0-8021-1839-4. Gladwell, Malcolm (2005). Blink. New York: Little, Brown and Co. ISBN 978-0-316-17232-5. Gödel, Kurt (1951). Some basic theorems on the foundations of mathematics and their implications. Gibbs Lecture. In Feferman, Solomon, ed. (1995). Kurt Gödel: Collected Works, Vol. III: Unpublished Essays and Lectures. Oxford University Press. pp. 304–23. ISBN 978-0-19-514722-3. Haugeland, John (1985). Artificial Intelligence: The Very Idea. Cambridge, Mass.: MIT Press. ISBN 978-0-262-08153-5. Hawkins, Jeff; Blakeslee, Sandra (2005). On Intelligence. New York, NY: Owl Books. ISBN 978-0-8050-7853-4. Henderson, Mark (24 April 2007). "Human rights for robots? We're getting carried away". The Times Online. London. Archived from the original on 31 May 2014. Retrieved 31 May 2014. Hernandez-Orallo, Jose (2000). "Beyond the Turing Test". Journal of Logic, Language and Information. 9 (4): 447–466. doi:10.1023/A:1008367325700. S2CID 14481982. Hernandez-Orallo, J.; Dowe, D. L. (2010). "Measuring Universal Intelligence: Towards an Anytime Intelligence Test". Artificial Intelligence. 174 (18): 1508–1539. CiteSeerX 10.1.1.295.9079. doi:10.1016/j.artint.2010.09.006. Hinton, G. E. (2007). "Learning multiple layers of representation". Trends in Cognitive Sciences. 11 (10): 428–434. doi:10.1016/j.tics.2007.09.004. PMID 17921042. S2CID 15066318. Hofstadter, Douglas (1979). Gödel, Escher, Bach: an Eternal Golden Braid. New York, NY: Vintage Books. ISBN 978-0-394-74502-2. Holland, John H. (1975). Adaptation in Natural and Artificial Systems. University of Michigan Press. ISBN 978-0-262-58111-0. Archived from the original on 26 July 2020. Retrieved 17 December 2019. Howe, J. (November 1994). "Artificial Intelligence at Edinburgh University: a Perspective". Archived from the original on 15 May 2007. Retrieved 30 August 2007. Hutter, M. (2012). "One Decade of Universal Artificial Intelligence". Theoretical Foundations of Artificial General Intelligence. Atlantis Thinking Machines. 4. pp. 67–88. CiteSeerX 10.1.1.228.8725. doi:10.2991/978-94-91216-62-6_5. ISBN 978-94-91216-61-9. S2CID 8888091. Kahneman, Daniel; Slovic, D.; Tversky, Amos (1982). Judgment under uncertainty: Heuristics and biases. Science. 185. New York: Cambridge University Press. pp. 1124–31. doi:10.1126/science.185.4157.1124. ISBN 978-0-521-28414-1. PMID 17835457. S2CID 143452957. Kaplan, Andreas; Haenlein, Michael (2019). "Siri, Siri in my Hand, who's the Fairest in the Land? On the Interpretations, Illustrations and Implications of Artificial Intelligence". Business Horizons. 62: 15–25. doi:10.1016/j.bushor.2018.08.004. Katz, Yarden (1 November 2012). "Noam Chomsky on Where Artificial Intelligence Went Wrong". The Atlantic. Archived from the original on 28 February 2019. Retrieved 26 October 2014. "Kismet". MIT Artificial Intelligence Laboratory, Humanoid Robotics Group. Archived from the original on 17 October 2014. Retrieved 25 October 2014. Koza, John R. (1992). Genetic Programming (On the Programming of Computers by Means of Natural Selection). MIT Press. Bibcode:1992gppc.book.....K. ISBN 978-0-262-11170-6. Kolata, G. (1982). "How can computers get common sense?". Science. 217 (4566): 1237–1238. Bibcode:1982Sci...217.1237K. doi:10.1126/science.217.4566.1237. PMID 17837639. Kumar, Gulshan; Kumar, Krishan (2012). "The Use of Artificial-Intelligence-Based Ensembles for Intrusion Detection: A Review". Applied Computational Intelligence and Soft Computing. 2012: 1–20. doi:10.1155/2012/850160. Kurzweil, Ray (1999). The Age of Spiritual Machines. Penguin Books. ISBN 978-0-670-88217-5. Kurzweil, Ray (2005). The Singularity is Near. Penguin Books. ISBN 978-0-670-03384-3. Lakoff, George; Núñez, Rafael E. (2000). Where Mathematics Comes From: How the Embodied Mind Brings Mathematics into Being. Basic Books. ISBN 978-0-465-03771-1. Langley, Pat (2011). "The changing science of machine learning". Machine Learning. 82 (3): 275–279. doi:10.1007/s10994-011-5242-y. Law, Diane (June 1994). Searle, Subsymbolic Functionalism and Synthetic Intelligence (Technical report). University of Texas at Austin. p. AI94-222. CiteSeerX 10.1.1.38.8384. Legg, Shane; Hutter, Marcus (15 June 2007). A Collection of Definitions of Intelligence (Technical report). IDSIA. arXiv:0706.3639. Bibcode:2007arXiv0706.3639L. 07-07. Lenat, Douglas; Guha, R. V. (1989). Building Large Knowledge-Based Systems. Addison-Wesley. ISBN 978-0-201-51752-1. Lighthill, James (1973). "Artificial Intelligence: A General Survey". Artificial Intelligence: a paper symposium. Science Research Council. Lucas, John (1961). "Minds, Machines and Gödel". In Anderson, A.R. (ed.). Minds and Machines. Archived from the original on 19 August 2007. Retrieved 30 August 2007. Lungarella, M.; Metta, G.; Pfeifer, R.; Sandini, G. (2003). "Developmental robotics: a survey". Connection Science. 15 (4): 151–190. CiteSeerX 10.1.1.83.7615. doi:10.1080/09540090310001655110. S2CID 1452734. Maker, Meg Houston (2006). "AI@50: AI Past, Present, Future". Dartmouth College. Archived from the original on 3 January 2007. Retrieved 16 October 2008. Markoff, John (16 February 2011). "Computer Wins on 'Jeopardy!': Trivial, It's Not". The New York Times. Archived from the original on 22 October 2014. Retrieved 25 October 2014. McCarthy, John; Minsky, Marvin; Rochester, Nathan; Shannon, Claude (1955). "A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence". Archived from the original on 26 August 2007. Retrieved 30 August 2007.. McCarthy, John; Hayes, P. J. (1969). "Some philosophical problems from the standpoint of artificial intelligence". Machine Intelligence. 4: 463–502. CiteSeerX 10.1.1.85.5082. Archived from the original on 10 August 2007. Retrieved 30 August 2007. McCarthy, John (12 November 2007). "What Is Artificial Intelligence?". Archived from the original on 18 November 2015. Minsky, Marvin (1967). Computation: Finite and Infinite Machines. Englewood Cliffs, N.J.: Prentice-Hall. ISBN 978-0-13-165449-5. Archived from the original on 26 July 2020. Retrieved 18 November 2019. Minsky, Marvin (2006). The Emotion Machine. New York, NY: Simon & Schusterl. ISBN 978-0-7432-7663-4. Moravec, Hans (1988). Mind Children. Harvard University Press. ISBN 978-0-674-57616-2. Archived from the original on 26 July 2020. Retrieved 18 November 2019. Norvig, Peter (25 June 2012). "On Chomsky and the Two Cultures of Statistical Learning". Peter Norvig. Archived from the original on 19 October 2014. NRC (United States National Research Council) (1999). "Developments in Artificial Intelligence". Funding a Revolution: Government Support for Computing Research. National Academy Press. Needham, Joseph (1986). Science and Civilization in China: Volume 2. Caves Books Ltd. Newell, Allen; Simon, H. A. (1976). "Computer Science as Empirical Inquiry: Symbols and Search". Communications of the ACM. 19 (3): 113–126. doi:10.1145/360018.360022.. Nilsson, Nils (1983). "Artificial Intelligence Prepares for 2001" (PDF). AI Magazine. 1 (1). Archived (PDF) from the original on 17 August 2020. Retrieved 22 August 2020. Presidential Address to the Association for the Advancement of Artificial Intelligence. O'Brien, James; Marakas, George (2011). Management Information Systems (10th ed.). McGraw-Hill/Irwin. ISBN 978-0-07-337681-3. O'Connor, Kathleen Malone (1994). "The alchemical creation of life (takwin) and other concepts of Genesis in medieval Islam". University of Pennsylvania: 1–435. Archived from the original on 5 December 2019. Retrieved 27 August 2008. Oudeyer, P-Y. (2010). "On the impact of robotics in behavioral and cognitive sciences: from insect navigation to human cognitive development" (PDF). IEEE Transactions on Autonomous Mental Development. 2 (1): 2–16. doi:10.1109/tamd.2009.2039057. S2CID 6362217. Archived (PDF) from the original on 3 October 2018. Retrieved 4 June 2013. Penrose, Roger (1989). The Emperor's New Mind: Concerning Computer, Minds and The Laws of Physics. Oxford University Press. ISBN 978-0-19-851973-7. Poli, R.; Langdon, W. B.; McPhee, N. F. (2008). A Field Guide to Genetic Programming. Lulu.com. ISBN 978-1-4092-0073-4. Archived from the original on 8 August 2015. Retrieved 21 April 2008 – via gp-field-guide.org.uk. Rajani, Sandeep (2011). "Artificial Intelligence – Man or Machine" (PDF). International Journal of Information Technology and Knowledge Management. 4 (1): 173–176. Archived from the original (PDF) on 18 January 2013. Ronald, E. M. A. and Sipper, M. Intelligence is not enough: On the socialization of talking machines, Minds and Machines Archived 25 July 2020 at the Wayback Machine, vol. 11, no. 4, pp. 567–576, November 2001. Ronald, E. M. A. and Sipper, M. What use is a Turing chatterbox? Archived 25 July 2020 at the Wayback Machine, Communications of the ACM, vol. 43, no. 10, pp. 21–23, October 2000. "Science". August 1982. Archived from the original on 25 July 2020. Retrieved 16 February 2016. Searle, John (1980). "Minds, Brains and Programs" (PDF). Behavioral and Brain Sciences. 3 (3): 417–457. doi:10.1017/S0140525X00005756. Archived (PDF) from the original on 17 March 2019. Retrieved 22 August 2020. Searle, John (1999). Mind, language and society. New York, NY: Basic Books. ISBN 978-0-465-04521-1. OCLC 231867665. Archived from the original on 26 July 2020. Retrieved 22 August 2020. Shapiro, Stuart C. (1992). "Artificial Intelligence". In Shapiro, Stuart C. (ed.). Encyclopedia of Artificial Intelligence (PDF) (2nd ed.). New York: John Wiley. pp. 54–57. ISBN 978-0-471-50306-4. Archived (PDF) from the original on 1 February 2016. Retrieved 29 May 2009. Simon, H. A. (1965). The Shape of Automation for Men and Management. New York: Harper & Row. Archived from the original on 26 July 2020. Retrieved 18 November 2019. Skillings, Jonathan (3 July 2006). "Getting Machines to Think Like Us". cnet. Archived from the original on 16 November 2011. Retrieved 3 February 2011. Solomonoff, Ray (1956). An Inductive Inference Machine (PDF). Dartmouth Summer Research Conference on Artificial Intelligence. Archived (PDF) from the original on 26 April 2011. Retrieved 22 March 2011 – via std.com, pdf scanned copy of the original. Later published as Solomonoff, Ray (1957). "An Inductive Inference Machine". IRE Convention Record. Section on Information Theory, part 2. pp. 56–62. Tao, Jianhua; Tan, Tieniu (2005). Affective Computing and Intelligent Interaction. Affective Computing: A Review. LNCS 3784. Springer. pp. 981–995. doi:10.1007/11573548. Tecuci, Gheorghe (March–April 2012). "Artificial Intelligence". Wiley Interdisciplinary Reviews: Computational Statistics. 4 (2): 168–180. doi:10.1002/wics.200. Thro, Ellen (1993). Robotics: The Marriage of Computers and Machines. New York: Facts on File. ISBN 978-0-8160-2628-9. Archived from the original on 26 July 2020. Retrieved 22 August 2020. Turing, Alan (October 1950), "Computing Machinery and Intelligence", Mind, LIX (236): 433–460, doi:10.1093/mind/LIX.236.433, ISSN 0026-4423. van der Walt, Christiaan; Bernard, Etienne (2006). "Data characteristics that determine classifier performance" (PDF). Archived from the original (PDF) on 25 March 2009. Retrieved 5 August 2009. Vinge, Vernor (1993). "The Coming Technological Singularity: How to Survive in the Post-Human Era". Vision 21: Interdisciplinary Science and Engineering in the Era of Cyberspace: 11. Bibcode:1993vise.nasa...11V. Archived from the original on 1 January 2007. Retrieved 14 November 2011. Wason, P. C.; Shapiro, D. (1966). "Reasoning". In Foss, B. M. (ed.). New horizons in psychology. Harmondsworth: Penguin. Archived from the original on 26 July 2020. Retrieved 18 November 2019. Weizenbaum, Joseph (1976). Computer Power and Human Reason. San Francisco: W.H. Freeman & Company. ISBN 978-0-7167-0464-5. Weng, J.; McClelland; Pentland, A.; Sporns, O.; Stockman, I.; Sur, M.; Thelen, E. (2001). "Autonomous mental development by robots and animals" (PDF). Science. 291 (5504): 599–600. doi:10.1126/science.291.5504.599. PMID 11229402. S2CID 54131797. Archived (PDF) from the original on 4 September 2013. Retrieved 4 June 2013 – via msu.edu. "Applications of AI". www-formal.stanford.edu. Archived from the original on 28 August 2016. Retrieved 25 September 2016. Further reading DH Author, 'Why Are There Still So Many Jobs? The History and Future of Workplace Automation' (2015) 29(3) Journal of Economic Perspectives 3. Boden, Margaret, Mind As Machine, Oxford University Press, 2006. Cukier, Kenneth, "Ready for Robots? How to Think about the Future of AI", Foreign Affairs, vol. 98, no. 4 (July/August 2019), pp. 192–98. George Dyson, historian of computing, writes (in what might be called "Dyson's Law") that "Any system simple enough to be understandable will not be complicated enough to behave intelligently, while any system complicated enough to behave intelligently will be too complicated to understand." (p. 197.) Computer scientist Alex Pentland writes: "Current AI machine-learning algorithms are, at their core, dead simple stupid. They work, but they work by brute force." (p. 198.) Domingos, Pedro, "Our Digital Doubles: AI will serve our species, not control it", Scientific American, vol. 319, no. 3 (September 2018), pp. 88–93. Gopnik, Alison, "Making AI More Human: Artificial intelligence has staged a revival by starting to incorporate what we know about how children learn", Scientific American, vol. 316, no. 6 (June 2017), pp. 60–65. Johnston, John (2008) The Allure of Machinic Life: Cybernetics, Artificial Life, and the New AI, MIT Press. Koch, Christof, "Proust among the Machines", Scientific American, vol. 321, no. 6 (December 2019), pp. 46–49. Christof Koch doubts the possibility of "intelligent" machines attaining consciousness, because "[e]ven the most sophisticated brain simulations are unlikely to produce conscious feelings." (p. 48.) According to Koch, "Whether machines can become sentient [is important] for ethical reasons. If computers experience life through their own senses, they cease to be purely a means to an end determined by their usefulness to... humans. Per GNW [the Global Neuronal Workspace theory], they turn from mere objects into subjects... with a point of view.... Once computers' cognitive abilities rival those of humanity, their impulse to push for legal and political rights will become irresistible – the right not to be deleted, not to have their memories wiped clean, not to suffer pain and degradation. The alternative, embodied by IIT [Integrated Information Theory], is that computers will remain only supersophisticated machinery, ghostlike empty shells, devoid of what we value most: the feeling of life itself." (p. 49.) Marcus, Gary, "Am I Human?: Researchers need new ways to distinguish artificial intelligence from the natural kind", Scientific American, vol. 316, no. 3 (March 2017), pp. 58–63. A stumbling block to AI has been an incapacity for reliable disambiguation. An example is the "pronoun disambiguation problem": a machine has no way of determining to whom or what a pronoun in a sentence refers. (p. 61.) E McGaughey, 'Will Robots Automate Your Job Away? Full Employment, Basic Income, and Economic Democracy' (2018) SSRN, part 2(3) Archived 24 May 2018 at the Wayback Machine. George Musser, "Artificial Imagination: How machines could learn creativity and common sense, among other human qualities", Scientific American, vol. 320, no. 5 (May 2019), pp. 58–63. Myers, Courtney Boyd ed. (2009). "The AI Report" Archived 29 July 2017 at the Wayback Machine. Forbes June 2009 Raphael, Bertram (1976). The Thinking Computer. W.H.Freeman and Company. ISBN 978-0-7167-0723-3. Archived from the original on 26 July 2020. Retrieved 22 August 2020. Scharre, Paul, "Killer Apps: The Real Dangers of an AI Arms Race", Foreign Affairs, vol. 98, no. 3 (May/June 2019), pp. 135–44. "Today's AI technologies are powerful but unreliable. Rules-based systems cannot deal with circumstances their programmers did not anticipate. Learning systems are limited by the data on which they were trained. AI failures have already led to tragedy. Advanced autopilot features in cars, although they perform well in some circumstances, have driven cars without warning into trucks, concrete barriers, and parked cars. In the wrong situation, AI systems go from supersmart to superdumb in an instant. When an enemy is trying to manipulate and hack an AI system, the risks are even greater." (p. 140.) Serenko, Alexander (2010). "The development of an AI journal ranking based on the revealed preference approach" (PDF). Journal of Informetrics. 4 (4): 447–459. doi:10.1016/j.joi.2010.04.001. Archived (PDF) from the original on 4 October 2013. Retrieved 24 August 2013. Serenko, Alexander; Michael Dohan (2011). "Comparing the expert survey and citation impact journal ranking methods: Example from the field of Artificial Intelligence" (PDF). Journal of Informetrics. 5 (4): 629–649. doi:10.1016/j.joi.2011.06.002. Archived (PDF) from the original on 4 October 2013. Retrieved 12 September 2013. Sun, R. & Bookman, L. (eds.), Computational Architectures: Integrating Neural and Symbolic Processes. Kluwer Academic Publishers, Needham, MA. 1994. Tom Simonite (29 December 2014). "2014 in Computing: Breakthroughs in Artificial Intelligence". MIT Technology Review. Tooze, Adam, "Democracy and Its Discontents", The New York Review of Books, vol. LXVI, no. 10 (6 June 2019), pp. 52–53, 56–57. "Democracy has no clear answer for the mindless operation of bureaucratic and technological power. We may indeed be witnessing its extension in the form of artificial intelligence and robotics. Likewise, after decades of dire warning, the environmental problem remains fundamentally unaddressed.... Bureaucratic overreach and environmental catastrophe are precisely the kinds of slow-moving existential challenges that democracies deal with very badly.... Finally, there is the threat du jour: corporations and the technologies they promote." (pp. 56–57.)
ReverendBayes / YOLO11m Car Damage DetectorCustom YOLO11m model for detecting and classifying car body damage (99% shattered glass, 96% flat tire detection accuracy)—optimized for high-capacity inference and assistive use in inspection and service workflows like BMW pre-loaner inspections.
StochSS / StochssStochSS - Software as a Service for modeling, simulation, inference and analysis of biochemical models. https://live.stochss.se
microsoft / InnerEye GatewayThe InnerEye-Gateway is a Windows service that acts as a DICOM end point to run inference on https://github.com/microsoft/InnerEye-DeepLearning models.
