PhysBench

KO | EN

gitlite — search

#python #hacktoberfest #react #ai #go #golang #deep-learning #linux #nodejs #macos #cli #kubernetes #awesome-list #vue

(★ 10)

A large-scale training and benchmarking framework for rPPG.

PhysBench 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# PhysBench
A large-scale training and benchmarking framework for rPPG.

Full release coming soon.

## What's new? 
* The training and testing code, refactored using Keras and JAX, is fully just-in-time (JIT) compiled. All models support FP16 mixed precision acceleration and Distributed Data Parallel (DDP), and can be deployed in both single-node multi-GPU environments and multi-node multi-GPU environments (under testing).
* A more flexible data loader that offers sub-dataset filtering capabilities, asynchronous streaming loading at speeds up to 4GB/s, and achieves nearly 100% GPU utilization. 
* Optimized APIs facilitate the addition of custom models and the application of data augmentation algorithms during training.

## In progress......
* Large-scale self-supervised pre-training for rPPG, supporting clusters of tens to hundreds of A100/H100.
* Multi-task fine-tuning using rPPG models to predict additional signals such as blood pressure (BP) and blood oxygen (SpO2).
* Continuous replication and benchmarking of state-of-the-art (SOTA) end-to-end models.
* Development documentation and function comments for beginners.

## Related works
* [A Plug-and-Play Temporal Normalization Module for Robust Remote Photoplethysmography](https://github.com/KegangWangCCNU/TemporalNormalization/tree/main)

The complete code is being organized. If you wish to add new features, please feel free to submit an issue.

PhysBench

PhysBench 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

PhysBench

PhysBench 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README