Keras ImageDataGenerator用于多个输入和基于图像的目标输出 [英] Keras ImageDataGenerator for multiple inputs and image based target output

问题描述

我有一个模型，该模型将两个图像作为输入，并生成一个图像作为目标输出.

I have a model which takes two Images as inputs and generates a single image as a Target output.

我所有的训练图像数据都在以下子文件夹中:

All of my training image-data is in the following sub-folders:

• input1
• input2
• 目标

我可以在keras中使用ImageDataGenerator类和诸如flow_from_directory和model.fit_generator方法的方法来训练网络吗?

Can I use the ImageDataGenerator class and methods like flow_from_directory and model.fit_generator method in keras to train the network?

我该怎么做?因为我遇到的大多数示例都涉及单个输入和基于标签的目标输出.

How can I do this? since most examples I have come across deal with single input and a label-based target output.

就我而言，我有一个非分类目标输出数据和多个输入.

In my case, I have a non-categorical target output data and multiple inputs.

请提供帮助，因为您的建议可能会很有帮助.

Please help, as your suggestions can be really helpful.

推荐答案

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.

这篇关于Keras ImageDataGenerator用于多个输入和基于图像的目标输出的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！