Keras 模型输出是 float32 而不是 uint8... 尽管数据标签是 uint8

我正在训练一个模型来预测医学图像中的分割。在训练数据中,输入数据类型为:numpy.float64,真实标签类型为:numpy.uint8。问题是出于某种原因,我的模型产生了 numpy.float32 的输出类型。

http://img.mukewang.com/61bafb890001caa610800756.jpg

# Defining the model

segmenter = Model(input_img, segmenter(input_img))


# Training the model (type of train_ground is numpy.uint8)

segmenter_train = segmenter.fit(train_X, train_ground, batch_size=batch_size,epochs=epochs,verbose=1,validation_data=(valid_X, valid_ground))

型号定义:


def segmenter(input_img):

    #encoder

    #input = 28 x 28 x 1 (wide and thin)

    conv1 = Conv2D(32, (3, 3), activation='relu', padding='same')(input_img) #28 x 28 x 32

    conv1 = BatchNormalization()(conv1)

    conv1 = Conv2D(32, (3, 3), activation='relu', padding='same')(conv1)

    conv1 = BatchNormalization()(conv1)

    pool1 = MaxPooling2D(pool_size=(2, 2))(conv1) #14 x 14 x 32

    conv2 = Conv2D(64, (3, 3), activation='relu', padding='same')(pool1) #14 x 14 x 64

    conv2 = BatchNormalization()(conv2)

    conv2 = Conv2D(64, (3, 3), activation='relu', padding='same')(conv2)

    conv2 = BatchNormalization()(conv2)

    pool2 = MaxPooling2D(pool_size=(2, 2))(conv2) #7 x 7 x 64

    conv3 = Conv2D(128, (3, 3), activation='relu', padding='same')(pool2) #7 x 7 x 128 (small and thick)

    conv3 = BatchNormalization()(conv3)

    conv3 = Conv2D(128, (3, 3), activation='relu', padding='same')(conv3)

    conv3 = BatchNormalization()(conv3)


白猪掌柜的
浏览 251回答 1
1回答

跃然一笑

Sigmoid 返回一个实数最后一层恰好是 sigmoid 激活函数。它返回一个从 0 到 1 的实数,而不是一个整数。此外,重要的是误差度量,即正确答案与计算值之间的差异,是连续的而不是离散的,因为它是可微的,并且允许通过反向传播正确学习神经网络权重。只需转换和舍入为了训练网络,只需将真值标签转换为浮点值。一旦你训练了网络并想要使用它的输出,只需将它们四舍五入以将它们转换为整数 - sigmoid 激活非常适合于此。
打开App,查看更多内容
随时随地看视频慕课网APP

相关分类

Python