VCDB background dataset is too large. How can I modify the vcdb.pickle?

xwen99 / temporal_context_aggregation

Temporal Context Aggregation for Video Retrieval with Contrastive Learning, WACV 2021

https://arxiv.org/abs/2008.01334

Apache License 2.0

27 stars 3 forks source link

VCDB background dataset is too large. How can I modify the vcdb.pickle? #6

Open chenqz1998 opened 2 years ago

chenqz1998 commented 2 years ago

The background dataset is very large (~1.1TB). I have to use the core dataset without background dataset to train the model. How can I modify the annotation file (vcdb.pickle)?

xwen99 commented 2 years ago

Hi,

Training without the background dataset may cause a significant drop in performance. As stated on the website of VCDB, please drop a mail to the authors for the background dataset.

chenqz1998 commented 2 years ago

Thanks for your reply. I have dropped a mail to the VCDB authors but received no replies. Could you please share the extracted features of VCDB and FIVR-5K to facilitate reproducibility for this research? Many thanks again.