Python Deep Learning Projects

Resizing

In the Cropping section, we saw that the original image of shape (480*720*3) is cropped to shape (393*254*3). However, the VGG16 architecture accepts images of shape (224*224*3). Hence, we will define a function called image_resize() that does the resizing for us. It accepts the cropped image and the joint resulting from the image_cropping() function as input and returns the resized image and its joint coordinates:

def image_resize(image, joints, new_size = 224):    """Function resize cropped images"""    orig_h, orig_w = image.shape[:2]    joints[0::2] = joints[0::2] / float(orig_w) * new_size    joints[1::2] = joints[1::2] / float(orig_h) * new_size    image = cv.resize(image, (new_size, new_size), interpolation=cv.INTER_NEAREST) return image, ...