search

kahnchana / svt

Official repository for "Self-Supervised Video Transformer" (CVPR'22)
https://kahnchana.github.io/svt
MIT License
104 stars 22 forks source link

config.MODEL.TWO_STREAM #15

Closed kffeng closed 1 year ago

kffeng commented 1 year ago

Dear team Thank you for sharing such a good work. I am very interested in your work. I am a beginner. I have read your paper and code, but there is a parameter（config.MODEL.TWO_STREAM） in the code that I don't understand. Is it input optical flow? If it is set to True, how can it work together config.MODEL.TWO_TOKEN?

kahnchana commented 1 year ago

That is part of experimental code that did not give good results. We did not use that setting in our paper.

It was left as a direction for future research: how to incorporate optical flow into the same system.