如何在多图像输入中使用Conv2D？

Question

我想使用多个图像作为网络的输入。我想添加Conv2D图层，类似这样的：

from tensorflow.keras.layers import *

from tensorflow.keras.models import Sequential

model = Sequential([

    Input(shape=(1, 128, 128, 1)),

    Conv2D(32, 3),

    Flatten(),

])

但这段代码会引发错误：Input 0 of layer conv2d_40 is incompatible with the layer: expected ndim=4, found ndim=5. Full shape received: [None, 1, 128, 128, 1]

但下面的代码工作正常：

model = Sequential([

    Input(shape=(1, 512, 512, 1)),

    Dense(32),

    Flatten(),

])

我知道，我可以添加多个Input图层，但我想知道有没有办法做到这样？

我的意思是我想使用输入形状的数据[NUMBER_OF_IMAGES, WIDTH, HEIGHT, N_CHANNELS]

并且NUMBER_OF_IMAGES不是所有图像的数量。这是当前输入的金额

Answer 1

Conv2D需要 4D 输入，您无法更改它。我不太确定你想要完成什么，但你可以使用Conv3D：from tensorflow.keras.layers import *from tensorflow.keras.models import Sequentialimport tensorflow as tfmodel = Sequential([&nbsp; &nbsp; Input(shape=(None, 128, 128, 1)),&nbsp; &nbsp; Conv3D(32, kernel_size=(1, 3, 3)),&nbsp; &nbsp; Flatten()])multiple_images = tf.random.uniform((10, 10, 128, 128, 1), dtype=tf.float32)model(multiple_images)<tf.Tensor: shape=(10, 5080320), dtype=float32, numpy=array([[-0.26742983, -0.09689523, -0.12120364, ..., -0.02987139,&nbsp; &nbsp; &nbsp; &nbsp; &nbsp;0.05515741,&nbsp; 0.12026916],&nbsp; &nbsp; &nbsp; &nbsp;[-0.18898709,&nbsp; 0.12448274, -0.17439063, ...,&nbsp; 0.23424357,&nbsp; &nbsp; &nbsp; &nbsp; -0.06001307, -0.13852882],&nbsp; &nbsp; &nbsp; &nbsp;[-0.14464797,&nbsp; 0.26356792, -0.34748033, ...,&nbsp; 0.07819699,&nbsp; &nbsp; &nbsp; &nbsp; -0.11639086,&nbsp; 0.10701762],&nbsp; &nbsp; &nbsp; &nbsp;...,&nbsp; &nbsp; &nbsp; &nbsp;[-0.1536693 ,&nbsp; 0.13642962, -0.18564&nbsp; &nbsp;, ...,&nbsp; 0.07165999,&nbsp; &nbsp; &nbsp; &nbsp; -0.0173855 , -0.04348694],&nbsp; &nbsp; &nbsp; &nbsp;[-0.32320747,&nbsp; 0.09207243, -0.22274591, ...,&nbsp; 0.11940736,&nbsp; &nbsp; &nbsp; &nbsp; -0.02635285, -0.1140241 ],&nbsp; &nbsp; &nbsp; &nbsp;[-0.21126074, -0.00094431, -0.10933039, ...,&nbsp; 0.06002581,&nbsp; &nbsp; &nbsp; &nbsp; -0.09649743,&nbsp; 0.09335127]], dtype=float32)>