what kind of input image format should be?

[cs-heibao]

@MacwinWin
hi, I use different pictures do the test but get the same predictions, and I wonder whether the input I used is wrong, and the input image format as follow:

[MacwinWin]

In my code, I use the code in Deep-Model-Transfer/preprocessing/inception_preprocessing.py to convert the image format. You forget to normalized the image to [-1,1], so the network can not recognize it correctly.

[cs-heibao]

@MacwinWin
hi, I've tried as you said, but the result looks wrong, so could you give me the instruction of the rest steps after I get the image by using cv2.imread, thanks.