CMD stands for Condensed Movies Dataset, which consists of key scenes from more than 3K movies: each key scene is accompanied by a high-level semantic description of the scene, character face tracks, and movie metadata. The dataset is scalable, automatically obtained from YouTube, and anyone can download and use it for free.

Citation

@misc{bain2020condensed, title={Condensed Movies: Story Based Retrieval with Contextual Embeddings}, author={Max Bain and Arsha Nagrani and Andrew Brown and Andrew Zisserman}, year={2020}, eprint={2005.04207}, archivePrefix={arXiv}, primaryClass={cs.CV} }