search

ygtxr1997 / MSML

MSML: Enhancing Occlusion-Robustness by Multi-Scale Segmentation-Based Mask Learning for Face Recognition
6 stars 5 forks source link

About 'cvt_ms1m_masked.py' #8

Closed a26z15 closed 1 year ago

a26z15 commented 1 year ago

Thanks for sharing the codes! Could you please provide the 'cvt_ms1m_masked.py' with more details, thanks a lot!

ygtxr1997 commented 1 year ago

We have some introduction about how to use mask augmentation in https://github.com/ygtxr1997/MSML/tree/main/datasets#3-masked-face-dataset. Dose this link help to solve your issue?

a26z15 commented 1 year ago

There is no code in 'cvt_ms1m_masked.py'. Is the codes in 'cvt_ms1m_masked.py' basically the same as 'cvt_casia_webface_masked.py'? Except the size of the image? Thanks for answering!

ygtxr1997 commented 1 year ago

Using cvt_casia_webface_masked.py can also process the MS1M dataset. You may just need to change the root_folder and make sure the root_folder contains train.rec and train.idx files. Because our CASIA dataset and MS1M dataset were both downloaded from https://github.com/deepinsight/insightface/tree/master/recognition/datasets. They have the same dataset format.

a26z15 commented 1 year ago

I see. Thanks again！