JEDE: Universal Jersey Number Detector for Sports

This codebase contains code for the paper "JEDE: Universal Jersey Number Detector for Sports" published on IEEE TCSVT, 2022.

Installation

Dependencies

Tested on PyTorch 1.8.1.

conda install pytorch==1.8.0 torchvision==0.9.0 torchaudio==0.8.0 cudatoolkit=10.2 -c pytorch
pip install opencv-python

TextSpotter v3

pip install pyclipper shapely

Our code base and Detectron2

git clone https://github.com/liuhengyue/pgrcnn.git
cd pgrcnn
git submodule init
git submodule update
python -m pip install -e detectron2

See installation instructions for more details on installing detectron2.

Weights

The weights trained with all images across soccer and basketball videos can be found in the release.

Dataset Preparation

Prepare Jersey Number

Currently, the dataset is not released due to policies. But, it will be released in the future.

mkdir datasets/jnw
ln -s datasets/jnw detectron2/datasets/jnw

Prepare COCO (Optional)

wget http://images.cocodataset.org/zips/train2017.zip
wget http://images.cocodataset.org/zips/val2017.zip
wget http://images.cocodataset.org/annotations/annotations_trainval2017.zip

mkdir datasets/coco
unzip train2017.zip -d datasets/coco
unzip val2017.zip -d datasets/coco
unzip annotations_trainval2017.zip -d datasets/coco

rm train2017.zip val2017.zip annotations_trainval2017.zip

Prepare SVHN (Optional)

wget http://ufldl.stanford.edu/housenumbers/train.tar.gz
mkdir svhn
tar -xvzf train.tar.gz -C svhn
rm train.tar.gz

Citations

If you find our work helpful, please cite:

@article{liu2022jede,
  title={JEDE: Universal Jersey Number Detector for Sports},
  author={Liu, Hengyue and Bhanu, Bir},
  journal={IEEE Transactions on Circuits and Systems for Video Technology},
  volume={32},
  number={11},
  pages={7894--7909},
  year={2022},
  publisher={IEEE}
}