search

TreB1eN / InsightFace_Pytorch

Pytorch0.4.1 codes for InsightFace
MIT License
1.74k stars 423 forks source link

About the format of images: RGB or BGR? #141

Closed doubbblek closed 4 years ago

doubbblek commented 4 years ago

Hi, first of all, thanks for sharing this great repo! Currently, I am doing some experience with the provided pretrained weights.

Here is the questions: In infer_on_video, I see you did not transform BGR images (from cv2) to RGM format. But in the data pipe, it seems like that you used PIL to load the image and as far as I know, this data type should be RGB instead of BGR. So which one is the correct data type for your pretrained model?

Best Regards

ruiming46zrm commented 4 years ago

cv2_read == pil_read + blue_img