Part of the official repo for paper "JBF: An Enhanced Representation of Skeleton for Video-based Human Action Recognition"

1. Create environment:

   conda create -n sns python=3.9 -y
   conda activate sns
   

2. Install PyTorch (Change verions according to your CUDA version):

   conda install pytorch==2.0.1 torchvision==0.15.2 pytorch-cuda=11.8 -c pytorch -c nvidia
   

3. Install OpenMM:

   pip install -U openmim
   mim install mmengine "mmcv==2.1.0" "mmdet==3.2.0"
   

4. Install additional dependencies:

   pip install -r requirements.txt
   pip install -v -e .
   

5. (Optional, bug exists with certain CUDA versions) Install alternate correlation implementation from RAFT:

   cd mmpose/models/flownets/alt_cuda_corr
   python setup.py install
   cd ../../../..