Questions about MLPs and G2L.

[czhaneva]

Thank you very much for the work. I have some questions.
(1) local MLPs. Take Resnet50 as an example, the feature dim of the last stage is 2048, according to the paper and the code, the in_dim of the local mlps will be 2048 * 9 = 18432. So the learnable parameters is 18423 * 18432 = 339,738,624 ~ 340 M >> Resnet50 backbone (25.5 M), Is it possible to train such a network ? And is it really reasonable to use such a huge MLPs ? I open this issue just for discussion.
(2) G2L. I use this idea in other task, but I found both global and local streams could converge, but the g2l could not converge. I'd like to ask that have you met this situation and how to solve this?
Thank you again.

[xieenze]

Thanks for the interest.

(1) I am not sure whether the parameter you calculated is accurate. But I remembered that the training GPU memory does not increase a lot. So adding mlp here do not have obvious overhead.

(2) I didn't meet such problem. But the weight of loss and temperature in contrastive loss are important for the performance, these are some experience from my side. You can try to tune these parameters

[czhaneva]

Thank you for your reply. Would you send me an appendix?
My email is [email protected]