weakly supervised training?

[WeihongM]

Will you share the details of the weakly supervised part?

Thanks.

[Real-YeJ]

@WeihongM For more details about the weakly supervised part, pls refer to our paper.

[WeihongM]

Hello, thanks for your reply. Recently I continue to finish this part of the work. I have some questions which I forgot to ask you before.

1. If there are no detection result in the current single real images, do you will calculate loss in this image (just continue it)?
2. What is more, whether the historical detection result will be utilized to generate the new pseudo label or just override to use the new detection result as the pseudo label? @Real-YeJ @ying09 @Alwinator