Pretrained backbones for resnet_ws_mrrp and resnet_ws_v2

[vadimkantorov]

Hi @shenyunhang!

The OneDrive link published in README: https://1drv.ms/f/s!Am1oWgo9554dgRQ8RE1SRGvK7HW2 has only pretrained resnet_ws models

Are resnet_ws_mrrp / resnet_ws_v2 models available somewhere? The resnet_ws_mrrp backbone seems the best-performing in the paper.

Btw, is resnet_ws or resnet_ws_v2 better performing? (and what is the meaning of difference in pooling?)

Thanks!

[shenyunhang]

Both resnet_ws_mrrp and resnet_ws_v2 use the same weights as resnet_ws. The differences are some model structures.