I found that in the training process, the image is resized by skimage.transform.resize method. In infer stage, if the resized method is one of cv2 inter cubic, area, ..., the matting result is worse normally. I want to know is it normal? I am a fresher in DL
I found that in the training process, the image is resized by
skimage.transform.resizemethod. In infer stage, if the resized method is one of cv2 inter cubic, area, ..., the matting result is worse normally. I want to know is it normal? I am a fresher in DL