This is an Implement of openpose with lstm module using TensorFlow.
The original openpose project based on caffe is here.
Only basic python is used, so the code is easy to understand.
The Dataloader and Post-processing code is from tf-pose-estimation.
- Ubuntu 16.04
- Cuda 10.0
- CuDNN 7.6.5
- Anaconda
Clone the repo and install 3rd-party libraries.
$ git clone https://github.com/GuaiYiHu/tf-et-assist-opensource
$ cd tf-et-assist-opensource
$ conda create -n py37 python=3.7
$ conda activate py37
$ pip3 install -r requirements.txt
Setup COCO API
$ git clone https://github.com/cocodataset/cocoapi.git
$ cd cocoapi/PythonAPI/
$ make
$ python setup.py install
$ cd ../../
Build c++ library for post processing.
$ sudo apt-get install swig
$ cd pafprocess
$ swig -python -c++ pafprocess.i && python3 setup.py build_ext --inplace
Download data.zip and checkpoints.zip from the link below.
BaiduNetDisk: Link: https://pan.baidu.com/s/1G3lJAsM0g_IkhwydRhTheg Password: bn7x
Google Drive: Link: https://drive.google.com/drive/folders/12v_Zw6kGxNDqHsP8hRqBGyFGjNJdgLZT?usp=sharing
To run test on single video in "data/" directory, run exp_signal_processing_single.py. Specify --video to the single video in "data/" directory in exp_signal_processing_single.py.
- For example, run
python exp_signal_processing_single.py --video 'data/1/平举.mp4'
If the environment is correct, the result of 'data/1/平举.mp4' will be the same as the picture below.
To run test on all videos in "data/" directory, run exp_signal_processing.py. Specify --mode or -m to the folder includes checkpoint files in exp_signal_processing.py.
- To run on calculate mode, run
python exp_signal_processing.py --mode cal
- To run on data test mode, run
python exp_signal_processing.py --mode test
If the environment is correct, the result of test mode will be the same as the picture below.