search

lxa9867 / R2VOS

Robust Referring Video Object Segmentation with Cyclic Structural Consistency [ICCV 2023]
25 stars 1 forks source link

About text encoder #3

Closed zyn213 closed 9 months ago

zyn213 commented 9 months ago

Hi! May I ask why you chose Roberta as your text encoder? Why didn't you use the text encoder from CLIP or Bert? Thank you!

lxa9867 commented 9 months ago

Thanks for your interest. We select Roberta following the referformer for a fair comparison.

We did try clip text encoder while we didn’t find a significant performance improvement in our experiments.

zyn213 commented 9 months ago

I see. Thank you so much!