Question of the test.py?

[Lin-plax]

def load_image(img_path):
image = cv2.imread(img_path, 0)
#image = cv2.resize(image, (112,112)).astype('float32')
image = cv2.resize(image, (128,128)).astype('float32')
if image is None:
return None
image = np.dstack((image, np.fliplr(image)))
image = image.transpose((2, 0, 1))
image = image[:, np.newaxis, :, :]
image = image.astype(np.float32, copy=False)
image -= 127.5
image /= 127.5
return image

why the number is 127.5 but not 128?

[tumble-weed]

if range of values in uint8 is between 0-255, then using 127.5 as mean makes minimum to be -127.5 and maximum to 127.5. try it with 128, and it wont be equal

[mtalhaubaid]

@tumble-weed @Lin-plax @ronghuaiyang do you have requirements file to run this repo?

[tumble-weed]

sorry @mtalhaubaid am not a maintainer of the repo, and since i tried running it within an existing environment, the requirements will be cluttered with other libraries as well as conda installed libraries.