[英]Input pipeline for semantic image segmentation (3 labels) with keras (TensforFlow backend) using flow_from_directory()
I am using keras (TensorFlow backend) and I am trying to understand how to bring in my labels/masks for image segmentation (3 labels) using flow_from_directory. 我正在使用keras(TensorFlow后端),并且试图了解如何使用flow_from_directory引入用于图像分割的标签/遮罩(3个标签)。
The train_images have the dimensions (144, 144, 144) - grayscale, uint8. train_images的尺寸为(144、144、144)-灰度,uint8。 The corresponding label_images have the same dimensions but here the value 1 represents label 1, value 2 = label 2, value 3 = label 3 and the value 0 shows unlabeled pixels.
相应的label_images具有相同的尺寸,但此处的值1表示标签1,值2 =标签2,值3 =标签3,值0显示未标签的像素。
Since this is semantic segmentation, classifying each pixel in the image requires using a pixel-wise cross-entropy loss function. 由于这是语义分割,因此对图像中的每个像素进行分类需要使用逐像素交叉熵损失函数。 And as I have read in some posts, keras (or TensorFlow) requires that my label_image/mask is one hot coded.
正如我在某些帖子中所读到的那样,keras(或TensorFlow)要求我的label_image / mask是一个热门代码。 Therefore I expect my label_images to be an image with 3 channels, where each pixel will consist of a binary vector.
因此,我希望我的label_images是具有3个通道的图像,其中每个像素将由一个二进制矢量组成。 Example: [0, 1, 0].
示例:[0,1,0]。
How do I deal with the unlabeled pixels that are stored as 0? 如何处理存储为0的未标记像素? Should they be encoded as [0, 0, 0]?
是否应该将它们编码为[0,0,0]?
But the question I have where I fail to find an answer is: How do I reshape/one-hot encode my label_images correctly? 但是,我在哪里找不到答案的问题是:如何正确整形/热编码label_images? Is there a handy function in keras that lets me convert my image_labels?
keras中有一个方便的函数可以让我转换image_labels吗?
from keras.preprocessing.image import ImageDataGenerator
train_datagen = ImageDataGenerator(rescale=1. / 255)
label_datagen = ImageDataGenerator(rescale=1. / 255)
train_image_generator = train_datagen.flow_from_directory(
directory='/train_images',
target_size=(144, 144, 144),
color_mode='grayscale',
classes=None,
class_mode=None,
batch_size=4)
train_label_generator = label_datagen.flow_from_directory(
directory='/label_images',
target_size=(144, 144, 144),
color_mode='grayscale',
classes=None,
class_mode=None,
batch_size=4)
train_generator = zip(train_image_generator, train_label_generator)
Currently working on something very similar but with 10 classes. 目前正在从事非常相似的工作,但有10个课程。 Still not entirely there yet, but as to you question about built-in functions for keras, checkout:
尚不完全存在,但是对于您关于keras内置函数的问题,请结帐:
one_hot_array = keras.utils.to_categorical(array_of_label_data, nb_classes)
which creates a one-hot-vector of your mask/label data. 这将创建您的遮罩/标签数据的一站式矢量。 So for your case, the expect output for say 100 masks would be (100, H, W, 3), where 3 is equal to the number of classes you're working with.
因此,对于您的情况,假设100个蒙版的预期输出为(100,H,W,3),其中3等于您正在使用的类的数量。 What I'm not sure about is if you do or do not have a background in your mask, and also, how you're supposed to structure the folders for your data.
我不确定的是您的蒙版是否有背景,以及如何为数据构建文件夹。 Hope that helps though.
希望能有所帮助。
Also, your target_size
is off, that's referring to the dimensions of your images (eg height and width). 同样,您的
target_size
处于关闭状态,这是指图像的尺寸(例如,高度和宽度)。 There shouldn't be a third value. 不应有第三个值。
声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.