cavaface

cavaface copied to clipboard

Maxpool in shortcut is redundant

1

In the Bottleneck_IR_SE , the shortcut layer of Maxpool is always with kernel=1 and stride=1. So the layer is useless. Am I missing something?

liorkad3

Thanks for the great work. Please do you support partial fc?

Source82

Remove redundancy in metrics.py

2

xsacha

In the paper, the model size is about 4M, why is there such a big difference? The optimizer and other information were not saved when saving the model.The training code...

lalazzd

Thanks for this work. I want to use the pre-trained model of Mobilefacenet for fine-tuning. Where can I get the pre-trained model?

simochen

https://github.com/cavalleria/cavaface.pytorch/blob/e9b9bd8ee06de51649ee202712d673f8e64415e9/backbone/resattnet.py#L79 Is it intentional that the stage1 attention block uses addition for the out_trunk whereas the rest use multiplication? Other repositories that implement this method appear to use multiply here,...

xsacha

xianfenggao

The test result(megaface rankn and roc) of Circle loss is so bad.

10

The test result(megaface rankn and roc) of Circle loss is so bad. I trained it by efficient-b0. but when I test it on megaface. the result is so bad. Do...

ReverseSystem001

Thanks for sharing this great work. I was just trying out your pretrained models for inference and was surprised that you are actually sharing torch jitted models? It is very...

rafale77

When I test mageface and IJB, where are the final test results saved or displayed？

3

When I test mageface and IJB, where are the final test results saved or displayed？ My final output is like this： Total extract time cost: 163.052476m free(): invalid pointer run.sh:...

Youngboy12