Result about MSVD

[shufangxun]

Hi, Bain
I found the experiment setting about MSVD result is not clear, meanwhile MCQ claims a different version. so I want to ask the result and setting about MSVD.

[m-bain]

Hi, sorry I was looking over old code. I treated them separately. Every caption->video retrieval pair counts towards the metric.

I avoid the nuanced setting of video->multi-caption retrieval by only reporting t2v!

Honestly there should be some standardised code evaluate all these benchmarks (i think https://github.com/albanie/collaborative-experts does this well but a tad hard to follow)

[shufangxun]

Thank you for your reply ! By the way, whether the result about MSVD under the setting of finetune or zero-shot?