Awesome
MoCA: Motion-Conditioned Image Animation for Video Editing
Wilson Yan<sup>1,2</sup>, Andrew Brown<sup>1</sup>, Pieter Abbeel<sup>2</sup>, Rohit Girdhar<sup>1</sup>, Samaneh Azadi<sup>1</sup>
<sup>1</sup>GenAI, Meta, <sup>2</sup> UC Berkeley
VideoEdit Benchmarking Data
To download the VideoEdit evaluation dataset in a local directory as ${DIR}, run
sudo apt-get update && sudo apt-get install ffmpeg
pip install -r requirements.txt
bash dataset/download_process_all.sh ${DIR}
This dataset includes the motion editing examples we collected from YouTube-8M, as well as filtered examples (excluding human faces and hands due to legal concerns) from Loveu-tgve-2023 data and the Dreamix examples. Please make sure to cite them accordingly.
License
VideoEdit data is released under CC-BY-NC 4.0 license. See License for additional details.