This repository contains code and data related to submission to ACL ARR 2024 - DeVAn: Dense Video Annotation for Video-Language Models.
For more details on the dataset and example videos and annotations, refer to our website.
Please see data/devan_acl24_release_v1.jsonl.gz for V1 of release data.
V1 release contains 8170 videos, their corresponding youtube id, start/end timestamp, ground truth… See the full description on the dataset page:
https://huggingface.co/datasets/TK-21st/DeVAn.