Closed LimGeunTaekk closed 3 years ago
Hi @LimGuenTaek,
Thanks for your interest in our work. To train BMN on TSP features, we use the same trick of interpolation I explained here. The interpolation is mainly to avoid changing anything in the code of BMN and just swap the input features.
Cheers!
Hi, Thanks a lot for sharing your code and features. It helped me a lot in my work.
Actually ... I tried to replicate Action Proposal Generation using BMN framework with TSP feature.
TSP framework returns Video feature : (num_of_clips , 512)
When I looked it up, BMN used features in the dimension (100,400).
I'm curious how you got the features to train BMN for action proposal generation.
I looked at issue and wondered if we just need to interpolate.
(num_of_clips , 512) -> 100 , 512
I would be so grateful if you could tell me the details.
Thanks for Your Paper , Work and Code