将 3D 数组调整为 2D

Question

I have some code that looks like this

    number_of_pairs = int(len(path_list) / 2)
    pairs_of_images = [np.zeros(
        (number_of_pairs, self.image_height, self.image_height, 1)) for i in range(2)]
    labels = np.zeros((number_of_pairs, 1))
    size = 105,105
    for pair in range(number_of_pairs):
        image = Image.open(path_list[pair * 2])
        image = image.resize((105,105))
        image = np.asarray(image).astype(np.float64)
        print("before resize is{}".format(image))

        pairs_of_images[0][pair, :, :, 0] = image

However, I'm getting an error where

pairs_of_images[1][pair, :, :, 0] = image
ValueError: could not broadcast input array from shape (105,105,4) into shape (105,105)

Is there a way to get rid of the 3rd dimension of the array?

Answer 1

Replace

image = Image.open(path_list[pair * 2])

with

image = Image.open(path_list[pair * 2]).convert('L')

'L' mode turns an image into a single-channel image.