Installation

Clone and enter this repository:

git clone https://github.com/dvl-tum/center-group
cd center-group

(Recommended) Create a conda environment for this repo:

conda create -n centergroup python=3.8.5 -y
conda activate centergroup

Install PyTorch:
```
conda install pytorch==1.8.1 torchvision==0.9.1 cudatoolkit=10.1 -c pytorch
```
These are the versions we used. Newer ones may also work. Make sure your system's CUDA version matches the one you specify for cudatoolkit here.
Install mm-cv:
```
pip install mmcv-full==1.3.10 -f https://download.openmmlab.com/mmcv/dist/cu101/torch1.8.1/index.html
```
Make sure that you replace cu101 for the CUDA version that you used to install PyTorch, if you used a different one.

Install this version of mmpose, and all of its requirements. To do so, you can run the following:

git clone [email protected]:open-mmlab/mmpose.git
cd mmpose
git checkout 65eb45b
pip install -r requirements.txt
python setup.py develop

Install Pytorch Lightning (needed for training):
```
pip install pytorch-lightning
```
Data preparation. Please follow the instructions here to download the COCO Keypoints and CrowdPose datasets in a directory of your choice. You don't need to download the person detection results. Please download the datasets under $CENTERGROUP_ROOT/data, or update the entry data_root in the config files if you store them somewhere else.
Download our trained models in $CENTERGROUP_ROOT/models, you can find them here

Provide feedback