Awesome

⚠️ Please note that all materials (codes, dataset, etc.) are available for non-commercial use only. ⚠️ Please refrain from using the materials for any commercial purposes in any form or capacity.

Neuro-Symbolic Action Quality Assessment (NS-AQA)

🏆 CVPR 2024 CVSports Best Paper Award

This repository contains the Python code implementation of NS-AQA for platform diving.

📝 Technical Paper: link

🤗 Huggingface Demo: link

Overview

We propose a neuro-symbolic paradigm for AQA.

Run NS-AQA

Score Report is saved as an HTML file at "./output/"

Run NS-AQA on a single dive clip.

python nsaqa.py path/to/video.mp4

Run NS-AQA on a single dive from the FineDiving Dataset. Each dive in the dataset has an instance id (x, y).

python nsaqa_finediving.py x y

# e.g. if the instance id is ('01', 1)
python nsaqa_finediving.py 01 1

Please consider citing our work:

@inproceedings{okamoto2024hierarchical,
  title={Hierarchical NeuroSymbolic Approach for Comprehensive and Explainable Action Quality Assessment},
  author={Okamoto, Lauren and Parmar, Paritosh},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={3204--3213},
  year={2024}
}