demo地址：https://github.com/keras-team/keras/blob/2.3.0/examples/conv_lstm.py
import numpy as np def generate_movies(n_samples=1200, n_frames=15): row = 80 col = 80 noisy__movies = np.zeros((n_samples, n_frames, row, col, 1), dtype=np.float) shifted__movies = np.zeros((n_samples, n_frames, row, col, 1), dtype=np.float) for i in range(n_samples): # Add 3 to 7 moving squares n = np.random.randint(3, 8) for j in range(n): # Initial position xstart = np.random.randint(20, 60) ystart = np.random.randint(20, 60) # Direction of motion directionx = np.random.randint(0, 3) - 1 directiony = np.random.randint(0, 3) - 1 # Size of the square w = np.random.randint(2, 4) for t in range(n_frames): x_shift = xstart + directionx * t y_shift = ystart + directiony * t noisy__movies[i, t, x_shift - w: x_shift + w, y_shift - w: y_shift + w, 0] += 1 # Make it more robust by adding noise. # The idea is that if during inference, # the value of the pixel is not exactly one, # we need to train the network to be robust and still # consider it as a pixel belonging to a square. # if np.random.randint(0, 2): # noise_f = (-1) ** np.random.randint(0, 2) # noisy_movies[i, t, x_shift - w - 1: x_shift + w + 1, y_shift - w - 1: y_shift + w + 1, 0] += noise_f * 0.1 # Shift the ground truth by 1 x_shift = xstart + directionx * (t + 1) y_shift = ystart + directiony * (t + 1) shifted__movies[i, t, x_shift - w: x_shift + w, y_shift - w: y_shift + w, 0] += 1 # Cut to a 40x40 window noisy_movies = noisy__movies[::, ::, 20:60, 20:60, ::] shifted_movies = shifted__movies[::, ::, 20:60, 20:60, ::] noisy_movies[noisy_movies >= 1] = 1 shifted_movies[shifted_movies >= 1] = 1 return noisy_movies, shifted_movies if __name__ == '__main__': noisy_movies, shifted_movies = generate_movies(n_samples=120) any_group_index = 0 # 对于每一组(15帧)图片, 不加噪声的情况下, noisy_movies的1:15帧的图片和shifted_movies的0:14帧图片一模一样 frames0 = noisy_movies[any_group_index][1:15] frames1 = shifted_movies[any_group_index][0:14] print((frames0 == frames1).all())