Existing action quality assessment (AQA) methods often require a large number of label annotations for fully supervised learning, which are laborious and expensive. In practice, the labeled data are difficult to obtain because the AQA annotation process requires domain-specific expertise. In this paper, we propose a novel semi-supervised method, which can be utilized for better assessment of the AQA task by exploiting a large amount of unlabeled data and a small portion of labeled data. Differing from the traditional teacher-student network, we propose a teacher-reference-student architecture to learn both unlabeled and labeled data, where the teacher network and the reference network are used to generate pseudo-labels for unlabeled data to supervise the student network.
2025-05💾 We released our code.2024-10🚀 Our paper accepted by ECCV 2024.
We utilize the Spearman’s rank correlation as an evaluation metric to assess the performance of our method under different labeled data.
- Prepare MTL-AQA dataset.
- Download the dataset from the link provided in this repo.
- Download the I3D backbone pretrained on Kinetics (this repo).
- Unzip it under the
dataset/folder.
- Dependencies
- python == 3.8.13
- torch== 1.12.1
- Create conda environment
conda create --name TRS python=3.8.13 source activate TRS pip install -r requirements.txt
Run the following code to start training.
python -u main.py --gpu 0 --exp 1 Our evaluation code is build upon USDL, CoRe. We acknowledge these team for their valuable contributions to the field of action quality assessment.
If you find this project useful for your research, please use the following BibTeX entry.
@inproceedings{yun2025semi,
title={Semi-Supervised Teacher-Reference-Student Architecture for Action Quality Assessment},
author={Yun, Wulian and Qi, Mengshi and Peng, Fei and Ma, Huadong},
booktitle={European Conference on Computer Vision},
pages={161--178},
year={2025},
organization={Springer}
}