This the repo for the paper "Foul prediction with estimated poses from soccer broadcast video" (Fang et el., 2023) This project aims to predict fouls in soccer matches using estimated poses from broadcast video footage. It utilizes advanced computer vision techniques to analyze video data and identify potential foul scenarios.
Recent advances in computer vision have made significant progress in tracking and pose estimation of sports players. However, there have been fewer studies on behavior prediction with pose estimation in sports, in particular, the prediction of soccer fouls is challenging because of the smaller image size of each player and of difficulty in the usage of e.g., the ball and pose information. In our research, we introduce an innovative deep learning approach for anticipating soccer fouls. This method integrates video data, bounding box positions, image details, and pose information by curating a novel soccer foul dataset. Our model utilizes a combination of convolutional and recurrent neural networks (CNNs and RNNs) to effectively merge information from these four modalities. The experimental results show that our full model outperformed the ablated models, and all of the RNN modules, bounding box position and image, and estimated pose were useful for the foul prediction. Our findings have important implications for a deeper understanding of foul play in soccer and provide a valuable reference for future research and practice in this area.
dataset: https://drive.google.com/file/d/1BFn6xIB1yodNBf29o3lb_OxjcWVLZf9n/view?usp=sharing
python futurefoul.py
[1] Giancola, S. et al. IEEE Conference, 1711–1721, 2018
[2] Zhang, et al. IEEE/CVF Conference , 889--898, 2019
[3] Giancola, S. et al. IEEE Conference, 4490–4499, 2021
[4] Scott, A. et al. IEEE/CVF Conference , 3569–3579, 2022
[5] Cioppa, A. et al. IEEE/CVF Conference , 3491–3502, 2022
...