trunpm/tpt-for-videoqa

tpt-for-videoqa's Issues

Video feature extraction model

Hello, what model is used in the second step to extract features， can you provide some information

When will the code released again?

Hello, I found this repository about 2 months ago, but re-opened with code all gone. I wonder whether the code will be released again and when. Thanks for your reply.

Temporal Pyramid

Thanks for you great works! In your code, the appearance feature dimension is (batch_size, levels, 16, 2048). If the level is setting as n, the total segments is , is it meaning the dimension levels is ? Did you compare the condition about random sample the same number of segments as temporal pyramid.
Thanks.