Skip to content

What and How Well You Performed? A Multitask Learning Approach to Action Quality Assessment [CVPR 2019]

Notifications You must be signed in to change notification settings

ParitoshParmar/MTL-AQA

Repository files navigation

MTL-AQA

What and How Well You Performed? A Multitask Learning Approach to Action Quality Assessment

🚀 Also Check Out Our New Approach! 🚀

Oct 2024: We have developed a new approach, NeuroSymbolic AQA, that builds upon this approach, but also analyses and scores using Professional Rules-based programs. It is Comprehensive and Explainable AQA which can generate Full Performance Reports for Actionable Insights!!! We encourage you to checkout [Code, Rules-based Programs, Dataset] [Demo] [Full Paper]

You are welcome to continue using this project, as it will still be maintained alongside the new approach!

MTL-AQA Concept:

diving_video

mtl_net

This repository contains MTL-AQA dataset + code introduced in the above paper. If you find this dataset or code useful, please consider citing:

@inproceedings{mtlaqa,
  title={What and How Well You Performed? A Multitask Learning Approach to Action Quality Assessment},
  author={Parmar, Paritosh and Tran Morris, Brendan},
  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
  pages={304--313},
  year={2019}
}

Check out our other relevant works:

Fine-grained Exercise Action Quality Assessment: Self-Supervised Pose-Motion Contrastive Approaches for Fine-grained Action Quality Assessment (can be used for Diving as well!) + Fitness-AQA dataset

*** Want to know the score of a Dive at the ongoing Olympics, even before the judges' decision? Try out our AI Olympics Judge ***