search

rese1f / MovieChat

[CVPR 2024] MovieChat: From Dense Token to Sparse Memory for Long Video Understanding
https://rese1f.github.io/MovieChat/
BSD 3-Clause "New" or "Revised" License
534 stars 41 forks source link

how do you build this moviechat-1k #45

Open sunwhw opened 8 months ago

sunwhw commented 8 months ago

Hi, Thanks for your great works! I have known the distribution of question type, but I still wanna know do you have any system guidelines or your own idea when designing question or do you just set different questions for different videos? And are the answers set manually?

image
Espere-1119-Song commented 8 months ago

We just set differen questions for different videos manually. And there exists another content-based categorization for the questions: image

sunwhw commented 8 months ago

oh, thanks! Can you give me the specific paper link? The latest version is https://arxiv.org/pdf/2307.16449, but I can't find the Table B.

Espere-1119-Song commented 8 months ago

Sorry, I can't give you the specific paper link now. Table B is from our rebuttal for CVPR 2024.

sunwhw commented 8 months ago

oo, Thanks a lot!

sunwhw commented 8 months ago

Will these updates be available in future releases?