131 skills found · Page 1 of 5
haomo-ai / MotionSeg3D[IROS 2022] Efficient Spatial-Temporal Information Fusion for LiDAR-Based 3D Moving Object Segmentation
malllabiisc / HyTEEMNLP 2018: HyTE: Hyperplane-based Temporally aware Knowledge Graph Embedding
apache / CtakesApache cTAKES is a Natural Language Processing (NLP) platform for clinical text.
clks-wzz / PRNet Depth GenerationA implementaion of depth generation based on [PRNet](https://github.com/YadiraF/PRNet), which was used in the paper ***Exploiting Temporal and Depth Information for Multi-frame Face Anti-spoofing***
rayat137 / Pose 3DExploiting temporal information for 3D pose estimation
jo-fu / TimeLineCuratorA visual timeline authoring tool that extracts temporal information from freeform text
LucipherDev / ComfyUI AniDocComfyUI Custom Nodes for "AniDoc: Animation Creation Made Easier". This approach automates line art video colorization using a novel model that aligns color information from references, ensures temporal consistency, and reduces manual effort in animation production.
Pose-Group / FAMI PoseThis is an official implementation of our CVPR 2022 ORAL paper "Temporal Feature Alignment and Mutual Information Maximization for Video-Based Human Pose Estimation" .
cumc-dbmi / CehrbertCEHR-BERT: Incorporating temporal information from structured EHR data to improve prediction tasks
radrumond / TimehetnetLearning complex time series forecasting models usually requires a large amount of data, as each model is trained from scratch for each task/data set. Leveraging learning experience with similar datasets is a well-established technique for classification problems called few-shot classification. However, existing approaches cannot be applied to time-series forecasting because i) multivariate time-series datasets have different channels and ii) forecasting is principally different from classification. In this paper we formalize the problem of few-shot forecasting of time-series with heterogeneous channels for the first time. Extending recent work on heterogeneous attributes in vector data, we develop a model composed of permutation-invariant deep set-blocks which incorporate a temporal embedding. We assemble the first meta-dataset of 40 multivariate time-series datasets and show through experiments that our model provides a good generalization, outperforming baselines carried over from simpler scenarios that either fail to learn across tasks or miss temporal information.
Telecommunication-Telemedia-Assessment / SITICommand-line tool for calculating Spatial Information / Temporal Information according to ITU-T P.910
HaydenFaulkner / TemporalEventAnnotatorA QT tool used to annotate temporal segments of a video with events which each have unique and customisable information.
koujan / Robotics Course ProjectHaze can cause poor visibility and loss of contrast in images and videos. In this article, we study the dehazing problem which can improve visibility and thus help in many computer vision applications. An extensive comparison of state of the art single image dehazing methods is done. One simple contrast enhancement method is used for dehazing. Structure- texture decomposition has been used in conjunction with this enhancement method to improve its performance in presence of synthetic noise. Methods which use a haze formation model and attempt at solving an ill-posed problem using computer vision priors are also investigated. The two priors studied are dark channel prior and the non-local prior. Both qualitative and quantitative comparisons for atmospheric and underwater images on all three methods provide a conclusive idea of which dehazing method performs better. All this knowledge has been extended to video dehazing. A video dehazing method which uses the spatial and temporal information in a video is studied in depth. An improved version of video dehazing is proposed in this article, which uses the spatial-temporal information fusion framework but does not suffer from some of its limitations. The new video dehazing method is shown to produce better results on test videos
AIS-Bonn / TemporalBallDetectionOfficial implementation of the paper: Utilizing Temporal Information in Deep Convolutional Network for Efficient Soccer Ball Detection and Tracking
JuliaEpi / MathEpiDeepLearningAwesome-spatial-temporal-data-mining-packages. Julia and Python resources on spatial and temporal data mining. Mathematical epidemiology as an application. Most about package information. Data Sources Links and Epidemic Repos are also included. Keep updating.
Intelligent-Computing-Lab-Panda / Exploring Temporal Information Dynamics In Spiking Neural NetworksPyTorch Implementation of Exploring Temporal Information Dynamics in Spiking Neural Networks (AAAI23)
SofanitAraya / CropPhenologyThis repository holds the code for an R package called CropPhenology. The package extracts phenological information of crops from multi temporal remote sensing vegetation index images
Hadryan / TFNet For Environmental Sound ClassificationLearning discriminative and robust time-frequency representations for environmental sound classification: Convolutional neural networks (CNN) are one of the best-performing neural network architectures for environmental sound classification (ESC). Recently, attention mechanisms have been used in CNN to capture the useful information from the audio signal for sound classification, especially for weakly labelled data where the timing information about the acoustic events is not available in the training data, apart from the availability of sound class labels. In these methods, however, the inherent time-frequency characteristics and variations are not explicitly exploited when obtaining the deep features. In this paper, we propose a new method, called time-frequency enhancement block (TFBlock), which temporal attention and frequency attention are employed to enhance the features from relevant frames and frequency bands. Compared with other attention mechanisms, in our method, parallel branches are constructed which allow the temporal and frequency features to be attended respectively in order to mitigate interference from the sections where no sound events happened in the acoustic environments. The experiments on three benchmark ESC datasets show that our method improves the classification performance and also exhibits robustness to noise.
pimlphm / Physics Informed Machine Learning Based On TCNA hybrid approach using physical information (PI) lightweight temporal convolutional neural networks (PI-TCN) for remaining useful life (RUL) prediction of bearings under stiffness degradation. It consists of three PI hybrid models: a) PI feature model (PIFM) - constructs physical information health indicators (PIHI) to increase the feature space; b) PI layer model (PILM) - encodes the physics governing equations in a hidden layer; c) PI layer-based loss model (PILLM) - designs PI conflicting losses, taking into account the integration of the physics input-output relationship module into the differences before and after the loss function. I have provided the original model and basic methodology here and welcome further optimisation of the structure and associated training methods. Interestingly, it is not the number of layers of physics knowledge that is more useful; the right structure for the right physics knowledge is the key to success. Similar to pure DL tuning, to design neural networks based on full physical knowledge is a direction that I am very interested in and would like to discuss with you.
hltfbk / E3C CorpusE3C is a freely available multilingual corpus (Italian, English, French, Spanish, and Basque) of semantically annotated clinical narratives to allow for the linguistic analysis, benchmarking, and training of information extraction systems. It consists of two types of annotations: (i) clinical entities: pathologies, symptoms, procedures, body parts, etc., according to standard clinical taxonomies (i.e. SNOMED-CT, ICD-10); and (ii) temporal information and factuality: events, time expressions, and temporal relations according to the THYME standard. The corpus is organised into three layers, with different purposes. Layer 1: about 25K tokens per language with full manual annotation of clinical entities, temporal information and factuality, for benchmarkingand linguistic analysis. Layer 2: 50-100K tokens per language with semi-automatic annotations of clinical entities, to be used to train baseline systems. Layer 3: about 1M tokens per language of non-annotated medical documents to be exploited by semi-supervised approaches. Researchers can use the benchmark training and test splits of our corpus to develop and test their own models. We trained several deep learning based models and provide baselines using the benchmark. Both the corpus and the built models will be available through the ELG platform.
