search

VIS-VAR / LGSC-for-FAS

Learning Generalized Spoof Cues for FaceAnti-spoofing
MIT License
225 stars 56 forks source link

patched input strategy #7

Closed mydlgit closed 4 years ago

mydlgit commented 4 years ago

In the paper, "we compare resizing the cropped faces to 224×224 with the bilinear interpolation and patching 224×224 the patches from the face images, and patched input strategy achieves better ACER metric." I don't understand what's 'patched input strategy', can you explan that? thx!!!

ZGSLZL commented 4 years ago

Hi，mydlgit, this strategy is to crop the 224x224 patches directly from the original face images, which means that there is no resize operation

VIS-VAR commented 4 years ago

Hi mydlgit, we have submitted new code with patched input strategy, please infer https://github.com/VIS-VAR/LGSC-for-FAS/pull/8