hightower

The result is all input samples are resized to cfg_width*cfg.height... This repo gives fine performance on lone text samples, but the pre-processing and post-processing module might need further development.

convert.py +line 25&&26 replace x = X_train[i] y = y_train[i] with x = X_test[i] y = y_test[i].

seems like the simdat is private lib of the auther: https://github.com/tammyyang/simdat.

I do have found the fact that the uploaded model pretrained from replay/3dmad database will kill too much live faces(dont remember the ratio but quite high) on other databases, though...

Hello, i guess not. The model works fine, at least on most samples(less than 90%). However, i dont think public anti-spoofing dataset is 'vivid' enough in practical scene, cause personally...

@CacheTechLtd https://github.com/cvtower/FASNet/FASNet.py see this file.

whats ur backend for keras there? I guess u should try keras version 2.0.9 or later?

hi @thinkInJava33 , I reproduced the performance from paper last year, and I will upload my modified version and trained models as well as tensorboard records later to my private...

> @cvtower thank you,i am wating for it,and i hope it won't be too long. hi @thinkInJava33 , https://github.com/cvtower/SeesawNet_pytorch/tree/master/cifar_test/78.21_igcv3_g2g2

> @1453042287 Hi, thanks for the advise. My current training seems working. > In my previous training, I set 'base' and 'loc' so on all in the trainable_scope, and it...