使用tfa.image.random_cutout應該完全滿足您的要求: import tensorflow as tfimport matplotlib.pyplot as pltimport tensorflow_addons as tfadef random_cut_out(images, labels): return tfa.image.random_cutout(images, (64, 64), constant_values = 1), labelsflowers = tf.keras.utils.get_file( 'flower_photos', 'https://storage.googleapis.com/download.tensorflow.org/example_images/flower_photos.tgz', untar=True)img_gen = tf.keras.preprocessing.image.ImageDataGenerator(rescale=1./255)ds = tf.data.Dataset.from_generator( lambda: img_gen.flow_from_directory(flowers, batch_size=32, shuffle=True), output_types=(tf.float32,