search

penincillin / DREAM

This is the public repository for our accepted CVPR 2018 paper "Pose-Robust Face Recognition via Deep Residual Equivariant Mapping"
http://mmlab.ie.cuhk.edu.hk/projects/DREAM/
BSD 2-Clause "Simplified" License
388 stars 98 forks source link

Are the given pretrained models derived from 'end2end' training or 'end2end + retrain' training? #34

Closed aman-captech closed 6 years ago

aman-captech commented 6 years ago

Are the pretrained models: cfp_res50_end2end and ijba_res18_end2end- just trained end2end or is the DREAM block retrained after end2end training?

penincillin commented 6 years ago

Both of the models come from end2end training

aman-captech commented 6 years ago

The novelty of the paper lies in the fact that the DREAM block is retrained on profile-frontal face pairs for pose robust face recognition. Is there any specific reason why you have not provided the retrained model and if possible can you provide it as it would be very helpful for a college project I am working on?

Thanks!

jolinlinlin commented 5 years ago

@penincillin The given models all come from end2end training.But you said "The strategy ‘end2end+retrain’ gives the best performance over all." So can you provide the retrained model or tell me how to retrain the DREAM block after end2end training? I can't find retrain code in your public code.Thank you very much!