一种可能性是使用class_mode=None(这样它们就不会返回任何目标)并使用shuffle=False(重要的).确保每个输入都使用相同的batch_size，并确保每个输入都位于不同的目录中，并且目标也位于不同的目录中，并且每个目录中的图像数量完全相同.

One possibility is to join three ImageDataGenerator into one, using class_mode=None (so they don't return any target), and using shuffle=False (important). Make sure you're using the same batch_size for each and make sure each input is in a different dir, and the targets also in a different dir, and that there are exactly the same number of images in each directory.

idg1 = ImageDataGenerator(...choose params...)
idg2 = ImageDataGenerator(...choose params...)
idg3 = ImageDataGenerator(...choose params...)

gen1 = idg1.flow_from_directory('input1_dir',
                                shuffle=False,
                                class_mode=None)
gen2 = idg2.flow_from_directory('input2_dir',
                                shuffle=False,
                                class_mode=None)
gen3 = idg3.flow_from_directory('target_dir',
                                shuffle=False,
                                class_mode=None)

创建自定义生成器:

class JoinedGen(tf.keras.utils.Sequence):
    def __init__(self, input_gen1, input_gen2, target_gen):
        self.gen1 = input_gen1
        self.gen2 = input_gen2
        self.gen3 = target_gen

        assert len(input_gen1) == len(input_gen2) == len(target_gen)

    def __len__(self):
        return len(self.gen1)

    def __getitem__(self, i):
        x1 = self.gen1[i]
        x2 = self.gen2[i]
        y = self.gen3[i]

        return [x1, x2], y

    def on_epoch_end(self):
        self.gen1.on_epoch_end()
        self.gen2.on_epoch_end()
        self.gen3.on_epoch_end()

使用此生成器进行训练:

Train with this generator:

my_gen = JoinedGen(gen1, gen2, gen3)
model.fit_generator(my_gen, ...)

或创建一个自定义生成器.上面显示了其所有结构.

Or create a custom generator. All the structure for it is shown above.