CheXNet-Keras icon indicating copy to clipboard operation
CheXNet-Keras copied to clipboard

preprocessing for cam

Open aguang1201 opened this issue 6 years ago • 2 comments

When you do cam preprocessing,the function transform_batch_images is not called.But the result of cam images is right.Why?Did you trained model without the function transform_batch_images?

aguang1201 avatar Aug 24 '18 05:08 aguang1201

I also noticed this difference. There is no normalization (with imagenet_mean and imagenet_std values) in cam.py because transform_batch_images is not called. Because of it test.py and cam.py return different predictions for same images. As I understand this normalization is a part of input pipeline during training, so this function should be called in cam.py, too.

dimitree54 avatar Sep 24 '18 19:09 dimitree54

yes, and the results look quite different when you add the normalization.

artembelopolsky avatar Feb 19 '21 14:02 artembelopolsky