/ASL

Code for CVPR'21 paper "Weakly Supervised Action Selection Learning in Video"

Primary LanguagePython

CVPR'21 Weakly Supervised Action Selection Learning in Video

[paper]

Authors: Jeremy Junwei Ma*, Satya Krishna Gorti*, Maksims Volkovs, Guangwei Yu

Introduction

This repository contains the implementation of ASL (Action Selection Learning) on the THUMOS-14 dataset.

Environment

The python code is developed and tested on the following environment:

  • Python 3.6
  • Pytorch 1.4.0

Experiments on THUMOS-14 dataset were run on a single NVIDIA TITAN V GPU with 12 GB GPU memory.

Dataset

Download the THUMOS-14 dataset here to DATA_PATH

Training

    bash ./scripts/train.sh DATA_PATH

Inference

    bash ./scripts/inference.sh DATA_PATH

Citation

If you find this code useful in your research, please cite the following paper:

@inproceedings{ma2021asl,
  title={Weakly Supervised Action Selection Learning in Video},
  author={Ma, Junwei and Gorti, Satya Krishna and Volkovs, Maksims and Yu, Guangwei},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  year={2021}
}