Optical flows associated with our work "We're Not Using Videos Effectively: An Updated Domain Adaptive Video Segmentation Baseline".
See the github for full instructions. But essentially just run
git lfs install
git clone https://huggingface.co/datasets/hoffman-lab/Unified-VideoDA-Generated-Flows