jingprism

I think there was a mistake with the code. If you input some large value like 200.0, it will produce real channel shifts like mentioned here under "Old Channel_Shift Images(cifar10)":...

Actually it used to work with that workaround in a previous version of keras, but not anymore. I looked through the code: https://github.com/keras-team/keras-preprocessing/blob/095aea9a029d52af618f3b1e8a4f93ef83f51d91/keras_preprocessing/image/image_data_generator.py#L872 https://github.com/keras-team/keras-preprocessing/blob/095aea9a029d52af618f3b1e8a4f93ef83f51d91/keras_preprocessing/image/affine_transformations.py#L159 And tried using the apply_channel_shift on...