howto100m icon indicating copy to clipboard operation
howto100m copied to clipboard

Published 20 hours ago • verified publisher icon antoine77340

Evaluation protocol

Open Alen-T opened this issue 4 years ago • 0 comments

Why “Learning a Text-Video Embedding from Incomplete and Heterogeneous Data” and “HowTo100M: Learning a Text-Video Embedding by Watching Hundred Million Narrated Video Clips” evaluation protocol different?

Is there a test set of 1k-A and 1k-B each representing 1000 randomly sampled text-video pairs?

I am very confused

Alen-T avatar Sep 24 '20 04:09 Alen-T