Unet分割的代码不含有数据集

人工智能

2.66MB

1 需要积分: 1

立即下载

资源介绍:

Unet分割的代码不含有数据集

U-Net: Convolutional Networks for Biomedical

Image Segmentation

Olaf Ronneberger, Philipp Fischer, and Thomas Brox

Computer Science Department and BIOSS Centre for Biological Signalling Studies,

University of Freiburg, Germany

ronneber@informatik.uni-freiburg.de,

WWW home page: http://lmb.informatik.uni-freiburg.de/

Abstract. There is large consent that successful training of deep net-

works requires many thousand annotated training samples. In this pa-

per, we present a network and training strategy that relies on the strong

use of data augmentation to use the available annotated samples more

eﬃciently. The architecture consists of a contracting path to capture

context and a symmetric expanding path that enables precise localiza-

tion. We show that such a network can be trained end-to-end from very

few images and outperforms the prior best method (a sliding-window

convolutional network) on the ISBI challenge for segmentation of neu-

ronal structures in electron microscopic stacks. Using the same net-

work trained on transmitted light microscopy images (phase contrast

and DIC) we won the ISBI cell tracking challenge 2015 in these cate-

gories by a large margin. Moreover, the network is fast. Segmentation

of a 512x512 image takes less than a second on a recent GPU. The full

implementation (based on Caﬀe) and the trained networks are available

at http://lmb.informatik.uni-freiburg.de/people/ronneber/u-net.

1 Introduction

In the last two years, deep convolutional networks have outperformed the state of

the art in many visual recognition tasks, e.g. [7,3]. While convolutional networks

have already existed for a long time [8], their success was limited due to the

size of the available training sets and the size of the considered networks. The

breakthrough by Krizhevsky et al. [7] was due to supervised training of a large

network with 8 layers and millions of parameters on the ImageNet dataset with

1 million training images. Since then, even larger and deeper networks have been

trained [12].

The typical use of convolutional networks is on classiﬁcation tasks, where

the output to an image is a single class label. However, in many visual tasks,

especially in biomedical image processing, the desired output should include

localization, i.e., a class label is supposed to be assigned to each pixel. More-

over, thousands of training images are usually beyond reach in biomedical tasks.

Hence, Ciresan et al. [1] trained a network in a sliding-window setup to predict

the class label of each pixel by providing a local region (patch) around that pixel

arXiv:1505.04597v1 [cs.CV] 18 May 2015

copy and crop

input

image

tile

output

segmentation

map

641

128

256

512

1024

max pool 2x2

up-conv 2x2

conv 3x3, ReLU

572 x 572

284²

128

256

512

570 x 570

568 x 568

282²

280²140²

138²

136²

68²

66²

64²

32²

28²

56²

54²

52²

512

104²

102²

100²

200²

30²

198²

196²

392 x 392

390 x 390

388 x 388

1024

512

256

128

64 2

conv 1x1

Fig. 1. U-net architecture (example for 32x32 pixels in the lowest resolution). Each blue

box corresponds to a multi-channel feature map. The number of channels is denoted

on top of the box. The x-y-size is provided at the lower left edge of the box. White

boxes represent copied feature maps. The arrows denote the diﬀerent operations.

as input. First, this network can localize. Secondly, the training data in terms

of patches is much larger than the number of training images. The resulting

network won the EM segmentation challenge at ISBI 2012 by a large margin.

Obviously, the strategy in Ciresan et al. [1] has two drawbacks. First, it

is quite slow because the network must be run separately for each patch, and

there is a lot of redundancy due to overlapping patches. Secondly, there is a

trade-oﬀ between localization accuracy and the use of context. Larger patches

require more max-pooling layers that reduce the localization accuracy, while

small patches allow the network to see only little context. More recent approaches

[11,4] proposed a classiﬁer output that takes into account the features from

multiple layers. Good localization and the use of context are possible at the

same time.

In this paper, we build upon a more elegant architecture, the so-called “fully

convolutional network” [9]. We modify and extend this architecture such that it

works with very few training images and yields more precise segmentations; see

Figure 1. The main idea in [9] is to supplement a usual contracting network by

successive layers, where pooling operators are replaced by upsampling operators.

Hence, these layers increase the resolution of the output. In order to localize, high

resolution features from the contracting path are combined with the upsampled

Fig. 2. Overlap-tile strategy for seamless segmentation of arbitrary large images (here

segmentation of neuronal structures in EM stacks). Prediction of the segmentation in

the yellow area, requires image data within the blue area as input. Missing input data

is extrapolated by mirroring

output. A successive convolution layer can then learn to assemble a more precise

output based on this information.

One important modiﬁcation in our architecture is that in the upsampling

part we have also a large number of feature channels, which allow the network

to propagate context information to higher resolution layers. As a consequence,

the expansive path is more or less symmetric to the contracting path, and yields

a u-shaped architecture. The network does not have any fully connected layers

and only uses the valid part of each convolution, i.e., the segmentation map only

contains the pixels, for which the full context is available in the input image.

This strategy allows the seamless segmentation of arbitrarily large images by an

overlap-tile strategy (see Figure 2). To predict the pixels in the border region

of the image, the missing context is extrapolated by mirroring the input image.

This tiling strategy is important to apply the network to large images, since

otherwise the resolution would be limited by the GPU memory.

As for our tasks there is very little training data available, we use excessive

data augmentation by applying elastic deformations to the available training im-

ages. This allows the network to learn invariance to such deformations, without

the need to see these transformations in the annotated image corpus. This is

particularly important in biomedical segmentation, since deformation used to

be the most common variation in tissue and realistic deformations can be simu-

lated eﬃciently. The value of data augmentation for learning invariance has been

shown in Dosovitskiy et al. [2] in the scope of unsupervised feature learning.

Another challenge in many cell segmentation tasks is the separation of touch-

ing objects of the same class; see Figure 3. To this end, we propose the use of

a weighted loss, where the separating background labels between touching cells

obtain a large weight in the loss function.

The resulting network is applicable to various biomedical segmentation prob-

lems. In this paper, we show results on the segmentation of neuronal structures

in EM stacks (an ongoing competition started at ISBI 2012), where we out-

资源文件列表:

unet-master - 副本.zip 大约有92个文件

unet-master - 副本/.idea/
unet-master - 副本/.idea/.gitignore 176B
unet-master - 副本/.idea/inspectionProfiles/
unet-master - 副本/.idea/inspectionProfiles/profiles_settings.xml 174B
unet-master - 副本/.idea/inspectionProfiles/Project_Default.xml 510B
unet-master - 副本/.idea/misc.xml 272B
unet-master - 副本/.idea/modules.xml 274B
unet-master - 副本/.idea/unet_42-new.iml 472B
unet-master - 副本/.idea/workspace.xml 7.83KB
unet-master - 副本/__pycache__/
unet-master - 副本/bmp转jpg.py 1KB
unet-master - 副本/bmp转png.py 1001B
unet-master - 副本/data/
unet-master - 副本/data/results/
unet-master - 副本/data/Test_Images/
unet-master - 副本/data/Test_Labels/
unet-master - 副本/data/Training_Images/
unet-master - 副本/data/Training_Labels/
unet-master - 副本/images/
unet-master - 副本/images/111/
unet-master - 副本/images/111/ISIC_0000000.jpg 48.79KB
unet-master - 副本/images/111/ISIC_0000000_res.png 4.76KB
unet-master - 副本/images/ISIC_0000000.jpg 48.79KB
unet-master - 副本/images/ISIC_0000000_res.png 4.72KB
unet-master - 副本/images/right.jpeg 26.57KB
unet-master - 副本/images/tmp/
unet-master - 副本/images/tmp/tmp_upload.jpeg 21.27KB
unet-master - 副本/images/UI/
unet-master - 副本/images/UI/logo.jpeg 33.37KB
unet-master - 副本/images/UI/lufei.png 215.7KB
unet-master - 副本/images/UI/right.jpeg 25.45KB
unet-master - 副本/images/UI/up.jpeg 27.88KB
unet-master - 副本/images/up.jpeg 21.27KB
unet-master - 副本/labelme2seg.py 1.02KB
unet-master - 副本/model/
unet-master - 副本/model/__init__.py
unet-master - 副本/model/__pycache__/
unet-master - 副本/model/__pycache__/__init__.cpython-37.pyc 141B
unet-master - 副本/model/__pycache__/__init__.cpython-38.pyc 161B
unet-master - 副本/model/__pycache__/unet_model.cpython-37.pyc 1.34KB
unet-master - 副本/model/__pycache__/unet_model.cpython-38.pyc 1.37KB
unet-master - 副本/model/__pycache__/unet_parts.cpython-37.pyc 2.79KB
unet-master - 副本/model/__pycache__/unet_parts.cpython-38.pyc 2.75KB
unet-master - 副本/model/unet_model.py 1.29KB
unet-master - 副本/model/unet_parts.py 3.39KB
unet-master - 副本/predict.py 1.72KB
unet-master - 副本/requirements.txt 143B
unet-master - 副本/results/
unet-master - 副本/results/confusion_matrix.csv 68B
unet-master - 副本/results/mIoU.png 15.56KB
unet-master - 副本/results/mPA.png 14.85KB
unet-master - 副本/results/Precision.png 14.7KB
unet-master - 副本/results/Recall.png 14.06KB
unet-master - 副本/test.py 4.17KB
unet-master - 副本/testdata/
unet-master - 副本/testdata/jsons/
unet-master - 副本/testdata/jsons/Case-1-U-1-1.json 65.34KB
unet-master - 副本/testdata/jsons/Case-2-U-2-2.json 77.86KB
unet-master - 副本/testdata/jsons/Case-2-U-2-3.json 89.63KB
unet-master - 副本/testdata/jsons/Case-3-U-5-0.json 83.71KB
unet-master - 副本/testdata/jsons/Case-3-U-5-2.json 83.12KB
unet-master - 副本/testdata/jsons/Case-3-U-5-3.json 80.91KB
unet-master - 副本/testdata/jsons/Case-4-U-8-0.json 84.39KB
unet-master - 副本/testdata/jsons/Case-4-U-8-1.json 85.46KB
unet-master - 副本/testdata/jsons/Case-5-U-10-0.json 78.91KB
unet-master - 副本/testdata/jsons/Case-5-U-10-1.json 82.62KB
unet-master - 副本/testdata/labels/
unet-master - 副本/testdata/labels/Case-1-U-1-1.png 3.18KB
unet-master - 副本/testdata/labels/Case-2-U-2-2.png 2.82KB
unet-master - 副本/testdata/labels/Case-2-U-2-3.png 3.25KB
unet-master - 副本/testdata/labels/Case-3-U-5-0.png 3.55KB
unet-master - 副本/testdata/labels/Case-3-U-5-2.png 3.59KB
unet-master - 副本/testdata/labels/Case-3-U-5-3.png 3.24KB
unet-master - 副本/testdata/labels/Case-4-U-8-0.png 3.02KB
unet-master - 副本/testdata/labels/Case-4-U-8-1.png 2.44KB
unet-master - 副本/testdata/labels/Case-5-U-10-0.png 3.47KB
unet-master - 副本/testdata/labels/Case-5-U-10-1.png 3.46KB
unet-master - 副本/train.py 2.46KB
unet-master - 副本/ui.py 8.04KB
unet-master - 副本/unet原文.pdf 1.57MB
unet-master - 副本/utils/
unet-master - 副本/utils/__pycache__/
unet-master - 副本/utils/__pycache__/dataset.cpython-37.pyc 1.81KB
unet-master - 副本/utils/__pycache__/dataset.cpython-38.pyc 1.84KB
unet-master - 副本/utils/__pycache__/utils_metrics.cpython-37.pyc 6.01KB
unet-master - 副本/utils/__pycache__/utils_metrics.cpython-38.pyc 6.35KB
unet-master - 副本/utils/data_remove_seg.py 730B
unet-master - 副本/utils/dataset.py 2.81KB
unet-master - 副本/utils/gen_split.py 1.57KB
unet-master - 副本/utils/label2png.py 1.41KB
unet-master - 副本/utils/utils_metrics.py 9.26KB
unet-master - 副本/切换镜像.txt 348B