Context-Based Prototype Learning for Weakly-Supervised Temporal Action Localization

Moayadeldin Hussain, Iker Gondra

Requirements and Dependencies

• GPU: GeForce GTX 1080 Ti
• CUDA: 11.8
• Python: 3.9.23
• PyTorch: 2.6.0+cu118
• NumPy: 1.24.4
• Pandas: 2.3.3

Required packages are provided in requirements.txt which can be installed using:

pip3 install -r requirements.txt

Data Preparation

We use the 2048-d features provided by MM 2021 paper: Cross-modal Consensus Network for Weakly Supervised Temporal Action Localization. You can get access of the dataset from Google Drive.

Testing

1. Download the pretrained model (47.6 Avg mAP on THUMOS'14 Benchmark) from Google Drive, and put it into "./download_ckpt/".

2. Run the following command:

python test.py --model_name CTXPL --dataset_name Thumos14reduced --path_dataset <C:\path_to_THUMOS14_dataset> --without_wandb

Citation

If you find our work useful in your research, please star 💫 the repo and cite the paper as follows:

@inproceedings{
hussain2026ctxpl,
title={Ctx{PL}: Context-based Prototype Learning for Weakly-Supervised Temporal Action Localization},
author={Moayadeldin Hussain and Iker Gondra},
booktitle={23rd Conference on Robots and Vision},
year={2026},
}

License

See MIT License

Acknowledgement

This repo contains modified codes from:

• ECCV2022-DELU: for implementation of the baseline model DELU.

This repo uses the features and annotations from:

• MM2021-CO2-Net: CO2-Net (MM2021).