A series of DeepMind Control environments with distractor backgrounds to test resistance to distractors.
python3 -m venv venv
source venv/bin/activate
pip3 install -r requirements.txt
You can download the Kinetics 400 dataset and grab the driving_car label from the train dataset. This is the dataset used by the original paper. Of course, you may use any other video dataset.
Some instructions for downloading the dataset can be found here: https://github.com/Showmax/kinetics-downloader. Just download it and drag-and-drop it to distractors.
This project is licensed under the CC-BY-NC 4.0, as found in the LICENSE file of this repo.
Original code by FacebookResearch for the "Learning Invariant Representations for Reinforcement Learning without Reconstruction" paper.
Original repo is here.
Formal attribution: "Learning Invariant Representations for Reinforcement Learning without Reconstruction" by https://github.com/facebookresearch/deep_bisim4control is licensed under CC-BY-NC 4.0.