search

facebookresearch / adaptive_teacher

This repo provides the source code for "Cross-Domain Adaptive Teacher for Object Detection".
Other
180 stars 35 forks source link

Can adaptive teacher use one-stage detector? #34

Closed huanmx closed 1 year ago

yujheli commented 1 year ago

I haven't tried it yet. I think if you append the adversarial loss on it maybe it would also work.

huanmx commented 1 year ago

Thanks for your reply! There is another question. Why not use discriminator to get loss_D_img_t in "supervise_target" branch rather than in "domain" branch. In other words, can we discard the "domain" branch and get loss_D_img_t in "supervise_target" branch and loss_D_img_s in "supervise" branch?

yujheli commented 1 year ago

@huanmx yeah, I think so. Since the loss is optimized at the same time for each iteration, that should also work.