search

ZhaoJ9014 / face.evoLVe

🔥🔥High-Performance Face Recognition Library on PaddlePaddle & PyTorch🔥🔥
MIT License
3.44k stars 759 forks source link

about data_pipe.py issues #113

Open konioyxgq opened 4 years ago

konioyxgq commented 4 years ago

In data_pipe.py ,load_mx_rec function header, img = mx.recordio.unpack_img(img_info)### img is BGR label = int(header.label) img = cv2.cvtColor(img, cv2.COLOR_RGB2BGR) ### img is RGB img = Image.fromarray(img) so train data is RGB. But ,in load_bin function img = mx.image.imdecode(_bin).asnumpy() ### img is RGB img = cv2.cvtColor(img, cv2.COLOR_RGB2BGR) ### img is BGR so test data is BGR.

Training data and test data formats are different.

konioyxgq commented 4 years ago

@ZhaoJ9014 Please! Help me! Did I get it wrong?

konioyxgq commented 4 years ago

Excuse me.could you give me a hand? @verasativa @clhne @insightcs @ZhaoJ9014

verasativa commented 4 years ago

I'm sorry, after my contribs to this repo, I got really worried that it may be used against oppressed minorities by totalitarian states, like Uyghurs by ccp. Not wanted to take the risk of helping that kind of evil, I decided not to involve in any biometrics stuff any more.

konioyxgq commented 4 years ago

Excuse me.