Thank for your amazing code.
The batch_size of comparative study is very large(eg.simclr'batch_size has about the hundreds), why your experiment batch_size equal to 16 can be carried out? Does the contrastive branch compare the proposals produced the RPN instead of the samples(different from Contrastive Learning)?
Thank for your amazing code. The batch_size of comparative study is very large(eg.simclr'batch_size has about the hundreds), why your experiment batch_size equal to 16 can be carried out? Does the contrastive branch compare the proposals produced the RPN instead of the samples(different from Contrastive Learning)?