RKTracker
:chart_with_upwards_trend: Real-time Keypoint Tracker using full/exhaustive-search motion estimation. Works with both live/camera streams and recorded video. Written in Python and C++, requires OpenCV, and optionally CUDA for GPGPU acceleration.
Install / Use
/learn @Breakthrough/RKTrackerREADME
RKTracker
Merging code from KeypointTrackerCpp repo and existing local cudaFSME code (GPGPU accelerated not publically released yet). Demonstration of existing GPGPU code (outlines what this repository will become):
This is the repository for Keypoint Tracker FSME, a program which implements full-search motion estimation (FSME) to track user-selected keypoints in real-time. A video file or camera/webcam stream can be opened and displayed, upon which you can click anywhere on the frame to add a keypoint at that location. In subsequent frames, the keypoints will continually be tracked and update their location (motion estimation) using a fully exhaustive sum-and-difference search to find the best match/position.
RKTracker is written in Python and C++, and requires OpenCV. If you have a CUDA-compatible GPU and the CUDA SDK, GPGPU acceleration can be enabled when compiling, allowing for a significant performance gain.
Licensed under BSD 2-Clause (see the LICENSE file for details).
Copyright (C) 2012-2016 Brandon Castellano. All rights reserved.
