Common functions and codes for image processing

numpy

jitter

# Randomly jitter the image a bit
ox, oy = np.random.randint(-max_jitter, max_jitter+1, 2)
X = np.roll(np.roll(X, ox, 1), oy, 2)
# Undo the jitter
X = np.roll(np.roll(X, -ox, 1), -oy, 2)        

• Roll array elements along a given axis.
  Elements that roll beyond the last position are re-introduced at the first.

>>> x = np.arange(10)
>>> np.roll(x, 2)
array([8, 9, 0, 1, 2, 3, 4, 5, 6, 7])

clip

X = np.clip(X, -SQUEEZENET_MEAN/SQUEEZENET_STD, (1.0 - SQUEEZENET_MEAN)/SQUEEZENET_STD)

• Given an interval, values outside the interval are clipped to the interval edges. For example, if an interval of [0, 1] is specified, values smaller than 0 become 0, and values larger than 1 become 1.

>>> a = np.arange(10)
>>> np.clip(a, 1, 8)
array([1, 1, 2, 3, 4, 5, 6, 7, 8, 8])

preprocess

def preprocess_image(img):
    """Preprocess an image for squeezenet.

    Subtracts the pixel mean and divides by the standard deviation.
    """
    return (img.astype(np.float32)/255.0 - SQUEEZENET_MEAN) / SQUEEZENET_STD

def deprocess_image(img, rescale=False):
    """Undo preprocessing on an image and convert back to uint8."""
    img = (img * SQUEEZENET_STD + SQUEEZENET_MEAN)
    if rescale:
        vmin, vmax = img.min(), img.max()
        img = (img - vmin) / (vmax - vmin)
    return np.clip(255 * img, 0.0, 255.0).astype(np.uint8)

ndarray[None]

equivalent to using numpy.newaxis