我定义了如下自编码器的CNN模型：
filters = (32, 16)X = Input(shape = (32, 32, 3))# encodefor f in filters: X = Conv2D(filters = f, kernel_size = (3, 3), activation = 'relu')(X) X = MaxPooling2D(pool_size = (2, 2), strides = (2, 2), padding = 'same')(X) X = BatchNormalization(axis = -1)(X)# decodefor f in filters[::-1]: X = Conv2D(filters = f, kernel_size = (3, 3), activation = 'relu')(X) X = UpSampling2D(size = (2, 2))(X) X = BatchNormalization(axis = -1)(X)模型摘要如下
Model: "functional_13"_________________________________________________________________Layer (type) Output Shape Param # =================================================================input_7 (InputLayer) [(None, 32, 32, 3)] 0 _________________________________________________________________conv2d_24 (Conv2D) (None, 30, 30, 32) 896 _________________________________________________________________max_pooling2d_12 (MaxPooling (None, 15, 15, 32) 0 _________________________________________________________________batch_normalization_24 (Batc (None, 15, 15, 32) 128 _________________________________________________________________conv2d_25 (Conv2D) (None, 13, 13, 16) 4624 _________________________________________________________________max_pooling2d_13 (MaxPooling (None, 7, 7, 16) 0 _________________________________________________________________batch_normalization_25 (Batc (None, 7, 7, 16) 64 _________________________________________________________________conv2d_26 (Conv2D) (None, 5, 5, 16) 2320 _________________________________________________________________up_sampling2d_12 (UpSampling (None, 10, 10, 16) 0 _________________________________________________________________batch_normalization_26 (Batc (None, 10, 10, 16) 64 _________________________________________________________________conv2d_27 (Conv2D) (None, 8, 8, 32) 4640 _________________________________________________________________up_sampling2d_13 (UpSampling (None, 16, 16, 32) 0 _________________________________________________________________batch_normalization_27 (Batc (None, 16, 16, 32) 128 =================================================================Total params: 12,864Trainable params: 12,672Non-trainable params: 192_________________________________________________________________由于输出图像的尺寸与输入图像不同，我得到了以下错误
InvalidArgumentError: Incompatible shapes: [128,32,32,3] vs. [128,16,16,32] [[node mean_squared_error/SquaredDifference (defined at <ipython-input-7-a9683921f595>:83) ]] [Op:__inference_train_function_21329]Function call stack:train_function因此无法计算损失函数。你能详细说明如何解决这个问题吗？
回答：
我建议你在卷积中使用padding='same'。还要注意不要用其他变量覆盖你的输入层。你还缺少一个最终输出层，其通道数应与输入图像的通道数相同
filters = (32, 16)inp = Input(shape = (32, 32, 3))# encodeX = inpfor f in filters: X = Conv2D(filters = f, kernel_size = (3, 3), padding = 'same', activation = 'relu')(X) X = MaxPooling2D(pool_size = (2, 2), strides = (2, 2), padding = 'same')(X) X = BatchNormalization(axis = -1)(X)# decodefor f in filters[::-1]: X = Conv2D(filters = f, kernel_size = (3, 3), padding = 'same', activation = 'relu')(X) X = UpSampling2D(size = (2, 2))(X) X = BatchNormalization(axis = -1)(X)out = Conv2D(filters = 3, kernel_size = (3, 3), padding = 'same')(X) model = Model(inp, out)现在的模型摘要如下
Layer (type) Output Shape Param # =================================================================input_9 (InputLayer) [(None, 32, 32, 3)] 0 _________________________________________________________________conv2d_22 (Conv2D) (None, 32, 32, 32) 896 _________________________________________________________________max_pooling2d_10 (MaxPooling (None, 16, 16, 32) 0 _________________________________________________________________batch_normalization_20 (Batc (None, 16, 16, 32) 128 _________________________________________________________________conv2d_23 (Conv2D) (None, 16, 16, 16) 4624 _________________________________________________________________max_pooling2d_11 (MaxPooling (None, 8, 8, 16) 0 _________________________________________________________________batch_normalization_21 (Batc (None, 8, 8, 16) 64 _________________________________________________________________conv2d_24 (Conv2D) (None, 8, 8, 16) 2320 _________________________________________________________________up_sampling2d_10 (UpSampling (None, 16, 16, 16) 0 _________________________________________________________________batch_normalization_22 (Batc (None, 16, 16, 16) 64 _________________________________________________________________conv2d_25 (Conv2D) (None, 16, 16, 32) 4640 _________________________________________________________________up_sampling2d_11 (UpSampling (None, 32, 32, 32) 0 _________________________________________________________________batch_normalization_23 (Batc (None, 32, 32, 32) 128 _________________________________________________________________conv2d_26 (Conv2D) (None, 32, 32, 3) 867 =================================================================Total params: 13,731Trainable params: 13,539Non-trainable params: 192_________________________________________________________________