时间戳

首页 > 解决方案 > 如何在 Keras 的自动编码器中拟合维度

keras - 如何在 Keras 的自动编码器中拟合维度

问题描述

我正在对 Mnist 图像数据（尺寸为 28*28）使用卷积自动编码器，这是我的代码

input_img = Input(shape=(28, 28, 1))

x = Convolution2D(16, (5, 5), activation='relu', padding='same')(input_img)
x = MaxPooling2D((2, 2), padding='same')(x)
x = Convolution2D(8, (3, 3), activation='relu', padding='same')(x)
x = MaxPooling2D((2, 2), padding='same')(x)
x = Convolution2D(8, (3, 3), activation='relu', padding='same')(x)
encoded = MaxPooling2D((2, 2), padding='same')(x)

x = Convolution2D(8, (3, 3), activation='relu', padding='same')(encoded)
x = UpSampling2D((2, 2))(x)
x = Convolution2D(8, (3, 3), activation='relu', padding='same')(x)
x = UpSampling2D((2, 2))(x)
x = Convolution2D(16, (5, 5), activation='relu', padding='same')(x)
x = UpSampling2D((2, 2))(x)


decoded = Convolution2D(1, (3, 3), activation='sigmoid', padding='same')(x)

我收到一条错误消息（每一层都有 padding ='same'）

ValueError: Error when checking target: expected conv2d_148 to have shape (32, 32, 1) but got array 
with shape (28, 28, 1)

这是我的模型摘要

层（类型）输出形状参数#

input_20 (InputLayer) (无, 28, 28, 1) 0

conv2d_142 (Conv2D) (无, 28, 28, 16) 416

max_pooling2d_64 (MaxPooling (无, 14, 14, 16) 0

conv2d_143 (Conv2D) (无, 14, 14, 8) 1160

max_pooling2d_65 (MaxPooling (无, 7, 7, 8) 0

conv2d_144 (Conv2D) (无, 7, 7, 8) 584

max_pooling2d_66 (MaxPooling (无, 4, 4, 8) 0

conv2d_145 (Conv2D) (无, 4, 4, 8) 584

up_sampling2d_64（上采样（无，8,8,8）0

conv2d_146 (Conv2D) (无, 8, 8, 8) 584

up_sampling2d_65（上采样（无，16,16,8）0

conv2d_147 (Conv2D) (无, 16, 16, 16) 3216

up_sampling2d_66（上采样（无，32、32、16）0

conv2d_148 (Conv2D) (无, 32, 32, 1) 145

总参数：6,689 可训练参数：6,689 不可训练参数：0

我知道如果我将第一层更改为 x = Convolution2D(16, (3, 3), activation='relu', padding='same')(input_img) 它可以工作，但我想使用 5*5 卷积。

它是怎么发生的？

标签： kerastf.kerasautoencoder

解决方案

您可以增加最后一个过滤器的大小来(5, 5)完成这项工作：

from tensorflow.keras.layers import *
from tensorflow.keras import Model, Input
import numpy as np

input_img = Input(shape=(28, 28, 1))

x = Conv2D(16, (5, 5), activation='relu', padding='same')(input_img)
x = MaxPooling2D((2, 2), padding='same')(x)
x = Conv2D(8, (3, 3), activation='relu', padding='same')(x)
x = MaxPooling2D((2, 2), padding='same')(x)
x = Conv2D(8, (3, 3), activation='relu', padding='same')(x)
encoded = MaxPooling2D((2, 2), padding='same')(x)

x = Conv2D(8, (3, 3), activation='relu', padding='same')(encoded)
x = UpSampling2D((2, 2))(x)
x = Conv2D(8, (3, 3), activation='relu', padding='same')(x)
x = UpSampling2D((2, 2))(x)
x = Conv2D(16, (5, 5), activation='relu', padding='same')(x)
x = UpSampling2D((2, 2))(x)

decoded = Conv2D(1, (5, 5), activation='sigmoid', padding='valid')(x)

auto = Model(input_img, decoded)

auto.build(input_shape=(1, 28, 28, 1))
auto(np.random.rand(1, 28, 28, 1)).shape

TensorShape([1, 28, 28, 1])

或者，使用tf.keras.Conv2DTranspose

推荐阅读