EARL

This repository provides a simple implementation of the work "Efficient Sparse Attacks on Videos using Reinforcement Learning"

Example:

Demo:

Here, we provide 5 video clips in folder "dataset_numpy", you can run the following script and observe the effect:

python un_attack_show.py   # untargeted attack

or

python T_attack_show.py    # targeted attack

Threat models:

The video classification model, please refer to project https://github.com/FenHua/action-recognition

You can train the recognition model with your own data, and use the video attack method to attack them.

Please cite:

It will be added soon!