This is the official repository for the following paper:

Chihiro Nakatani, Hiroaki Kawashima, Norimichi Ukita.
Human-in-the-loop Adaptation in Group Activity Feature Learning for Team Sports Video Retrieval.
Computer Vision and Image Understanding, vol.263, pp. 104577, 2026.
Project page: https://toyota-ti.ac.jp/Lab/Denshi/iim/ukita/selection/CVIU2026-GAFL.html

Our codes are based on https://github.com/JacobYuan7/DIN-Group-Activity-Recognition-Benchmark. and https://github.com/chihina/GAFL-CVPR2024. I deeply appreciate their efforts.

@article{NAKATANI2026104577,
title = {Human-in-the-loop adaptation in group activity feature learning for team sports video retrieval},
journal = {Computer Vision and Image Understanding},
volume = {263},
pages = {104577},
year = {2026},
author = {Chihiro Nakatani and Hiroaki Kawashima and Norimichi Ukita},
}

python 3.10.2
ROIAlign (https://github.com/longcw/RoIAlign.pytorch)

And you can use requirements.txt

pip install -r requirements.txt

You can download daatset from the following url.
These dataset are required to place in data/ in the repository as follows:

• Volleyball dataset (data/volleyball/videos)
  https://github.com/mostafa-saad/deep-activity-rec

• NBA dataset (data/basketball/videos)
  https://ruiyan1995.github.io/SAM.html

• Collective Activity dataset (data/collective)
  https://cvgl.stanford.edu/projects/collective/collectiveActivity.html

• You can change parameters of the model by editing the files located in scripts (e.g., scripts/run_multiple_volleyball.bash).

• Ours

bash scripts/run_multiple_volleyball.bash

• Ours

bash scripts/run_multiple_basketball.bash

• Ours

bash scripts/run_multiple_volleyball.bash