dataset link on Hugging dataset
https://huggingface.co/datasets/lixinhao/VideoEval
Arxiv link
https://arxiv.org/abs/1809.04094
Description of the dataset
FIVR-200K is a fine-grained video → video retrieval benchmark with 225,960 videos and 100 queries, annotated for multiple levels of video and incident similarity.
A reproducible FIVR-5K evaluation subset and annotations are available through the VideoEval Hugging Face repository. Original source: https://github.com/MKLab-ITI/FIVR-200K
Adding it for MOEB (#4842) to strengthen the currently thin video → video retrieval direction. Claimed and approved in #4842.
dataset link on Hugging dataset
https://huggingface.co/datasets/lixinhao/VideoEval
Arxiv link
https://arxiv.org/abs/1809.04094
Description of the dataset
FIVR-200K is a fine-grained video → video retrieval benchmark with 225,960 videos and 100 queries, annotated for multiple levels of video and incident similarity.
A reproducible FIVR-5K evaluation subset and annotations are available through the VideoEval Hugging Face repository. Original source: https://github.com/MKLab-ITI/FIVR-200K
Adding it for MOEB (#4842) to strengthen the currently thin video → video retrieval direction. Claimed and approved in #4842.