search

elijahcole / single-positive-multi-label

Multi-Label Learning from Single Positive Labels - CVPR 2021
https://arxiv.org/abs/2106.09708
MIT License
91 stars 18 forks source link

Asking for extra results #2

Closed SonDaoDuy closed 3 years ago

SonDaoDuy commented 3 years ago

Thank you for your great work! I feel that the online training procedure is quite independent to the loss L_Epr. Have you tried to combine the online training procedure with other losses? If it is convenient for you, can you provide some results on those combinations? And also I cannot find the supplementary file anywhere, can you release it?

elijahcole commented 3 years ago

Thanks for reading!

You're right that the online label estimation framework could be combined with different losses, and I agree that it would be interesting to try some different approaches. Due to other commitments, I probably won't be able to do too many of those experiments myself in the near future.

Regarding the supplementary material, I'm still finalizing a thing or two - sorry for the delay!

coldmanck commented 3 years ago

Hi @elijahcole how about the supplementary material? Thank you :-)

elijahcole commented 3 years ago

@coldmanck The supplementary material has been submitted to arxiv, and should be available Sunday night! Thanks for your patience.

coldmanck commented 3 years ago

@elijahcole thanks!!