图像语义分割

￡神魔★判官ぃ 2022-09-26 01:59 270阅读 0赞

https://handong1587.github.io/deep\_learning/2015/10/09/segmentation.html

http://blog.csdn.net/column/details/15893.html

pascal\_voc语义分割排行榜http://host.robots.ox.ac.uk:8080/leaderboard/displaylb.php?challengeid=11&compid=6

[图像分类，物体检测，语义分割，实例分割的联系和区别][Link 1]

[图像分割的性能测试指标小结][Link 2]

[从特斯拉到计算机视觉之「图像语义分割」][Link 3]

[关于图像语义分割的总结和感悟][Link 4]

[反卷积与语义分割][Link 5]

[全卷积网络（FCN）与图像分割][FCN]

[语义分割相关总结][Link 6]

## FCN ##

Fully Convolutional Networks for Semantic Segmentation

[深度学习入门:Fully Convolutional Networks][Fully Convolutional Networks]

[深度学习入门:Fully Convolutional Networks for Semantic Segmentation][Fully Convolutional Networks for Semantic Segmentation]

[Fully Convolutional Networks for Semantic Segmentation][Fully Convolutional Networks for Semantic Segmentation 1]

[深度学习（二十一）基于FCN的图像语义分割-CVPR 2015][FCN_-CVPR 2015]

[全卷积网络 FCN 详解][FCN 1]

1.  [FCN for Sematic Segmentation.pptx][]

[全卷积网络：从图像级理解到像素级理解][Link 7]

[Semantic Segmentation with FCN里的32x, 16x和8x][Semantic Segmentation with FCN_32x_ 16x_8x]

[语义分割--FCN 算法中的一些细节--特征怎么融合][--FCN _--]

## deeplab ##

项目主页http://liangchiehchen.com/projects/DeepLab.html

Semantic Image Segmentation with Deep Convolutional Nets and Fully Connected CRFs

[论文阅读笔记：图像分割方法deeplab以及Hole算法解析][deeplab_Hole]

[DeepLab论文解读][DeepLab]

[深度学习之DeepLab用于语义分割][DeepLab 1]

[\[Paper Reading\] DeepLab v1 & v2][Paper Reading_ DeepLab v1 _ v2]

[几分钟看完语义分割经典论文:DeepLab][DeepLab 2]

[DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs][DeepLab_ Semantic Image Segmentation with Deep Convolutional Nets_ Atrous Convolution_ and Fully Connected CRFs]

[CRF图像分割简介][CRF]

https://hsh.blog.ustc.edu.cn/wp-content/uploads/2015/09/Class-Segmentation-and-Object-Localization-with-Superpixel-Neighborhoods.pdf

[推荐一些用CRF做图像语义分割的资源][CRF 1]

[Conditional Random Fields as Recurrent Neural Networks][]

[深度学习（三十三）CRF as RNN语义分割-未完待续][CRF as RNN_-]

[CRF as RNN的原理及Caffe实现][CRF as RNN_Caffe]

[图像语义分割之FCN和CRF][FCN_CRF]

[CRF as RNN 代码解读][CRF as RNN]

1.  [Semantic Image Segmentation With Deep Convolutional Nets and Fully Connected CRFs.ppt][]

http://www.robots.ox.ac.uk/~szheng/CRFasRNN.html

## U-Net ##

[全卷机神经网络图像分割-keras实现][-keras]

## DeconvNet ##

[Learning Deconvolution Network for Semantic Segmentation][]

## MNC ##

[《Instance-aware Semantic Segmentation via Multi-task Network Cascades》(MNC)，用自己的数据运行][Instance-aware Semantic Segmentation via Multi-task Network Cascades_MNC]

[ＭＮＣ用自己的数据运行得到如下错误：IndexError: index 4 is out of bounds for axis 1 with size 4][IndexError_ index 4 is out of bounds for axis 1 with size 4]

[论文解读-<Instance-aware Semantic Segmentation via Multi-task Network Cascades>][-_Instance-aware Semantic Segmentation via Multi-task Network Cascades]

[图像分割“Instance-aware Semantic Segmentation via Multi-task Network Cascades”][Instance-aware Semantic Segmentation via Multi-task Network Cascades]

#     #

## ParseNet ##

[ParseNet论文解读][ParseNet]

[语义分割(semantic segmentation) 常用神经网络介绍对比-FCN SegNet U-net DeconvNet][semantic segmentation_ _-FCN SegNet U-net DeconvNet]

[多篇用DL做Semantic Segmentation的文章总结][DL_Semantic Segmentation]

## PSPNet ##

[深度学习之PSPnet用于语义分割][PSPnet]

[几分钟看完PSPnet：Pyramid Scene Parsing Network][PSPnet_Pyramid Scene Parsing Network]

[Pyramid Scene Parsing Network - arxiv - 1612.01105][]

[图像语义分割之特征整合和结构预测][Link 8]

ICNet

# 实时语义分割--ICNet for Real-Time Semantic Segmentation on High-Resolution Images #

https://github.com/hszhao/ICNet

http://blog.csdn.net/yaoliangsjtu/article/details/77620936

## 弱标注 ##

Weakly- and Semi-Supervised Learning of a Deep Convolutional Network for Semantic Image Segmentation

[CNN在基于弱监督学习的图像分割中的应用][CNN]

[论文笔记 | BoxSup: Exploiting Bounding Boxes to Supervise Convolutional Networks for Semantic Segmentati][_ BoxSup_ Exploiting Bounding Boxes to Supervise Convolutional Networks for Semantic Segmentati]

An Adversarial Regularisation for Semi-Supervised Training of Structured Output Neural Networks

[弱监督学习在语义分割中的进展与应用?][Link 9]

## 和GAN结合 ##

FCNs in the Wild: Pixel-level Adversarial and Constraint-based Adaptation

Semantic Segmentation using Adversarial Networks

[深度对抗学习在图像分割和超分辨率中的应用][Link 10]

[论文精读 Learning to Refine Object Segments][Learning to Refine Object Segments]

[基于Mutil-Scale CNN的图片语义分割、法向量估计-ICCV 2015][Mutil-Scale CNN_-ICCV 2015]

Simultaneous Detection and Segmentation (SDS)

Hypercolumn Object Segmentation

[\[Paper Reading\] Attention to Scale: Scale-aware Semantic Image Segmentation][Paper Reading_ Attention to Scale_ Scale-aware Semantic Image Segmentation]

[几分钟看完 Zifeng Wu的38层网络：Wider or Deeper: Revisiting the ResNet Model for Visual Recognition][Zifeng Wu_38_Wider or Deeper_ Revisiting the ResNet Model for Visual Recognition]

[几分钟看完RefineNet:RefineNet: Multi-Path Refinement Networks for High-Resolution Semantic Segmentation][RefineNet_RefineNet_ Multi-Path Refinement Networks for High-Resolution Semantic Segmentation]

[几分钟看完zifeng wu instance level segmentation：Bridging Category-level and Instance-level Semantic Image][zifeng wu instance level segmentation_Bridging Category-level and Instance-level Semantic Image]

[几分钟看完旷视孙剑最新大作-Large Kernel Matters —— Improve Semantic Segmentation by Global Convolutional Network][-Large Kernel Matters _ Improve Semantic Segmentation by Global Convolutional Network]

[论文笔记：Label Refinement Network for Coarse-to-Fine Semantic Segmentation][Label Refinement Network for Coarse-to-Fine Semantic Segmentation]

[Fully Convolutional Instance-aware Semantic Segmentation 论文笔记][Fully Convolutional Instance-aware Semantic Segmentation]

将CRF的refine替换为边缘检测方法

Semantic Image Segmentation with Task-Specific Edge Detection Using CNNs and a Discriminatively Trained Domain Transform

http://blog.csdn.net/zhangjunhit/article/category/6809642

[语义分割 - Semantic Segmentation Papers][- Semantic Segmentation Papers]

## FCN实践 ##

[Ubuntu下全卷积神经网络（FCN）框架的配置和基本使用][Ubuntu_FCN]

[FCN网络训练 终极版][FCN_]

[使用FCN做图像语义分割(实践篇)][FCN 2]

[net\_surgery中如何将全连接层转换成卷积层][net_surgery]

[在caffe上做FCN模型fine-tune的一些注意事项][caffe_FCN_fine-tune]

[FCN网络的训练——以SIFT-Flow 数据集为例][FCN_SIFT-Flow]

[FCN模型训练中遇到的困难][FCN 3]

[FCN网络训练 SIFTFLOW数据集][FCN_ SIFTFLOW]

[Tensoflow学习记录13--用深度学习来做图像分割Fully Convolutional Networks for Semantic Segmentation (FCNs)][Tensoflow_13--_Fully Convolutional Networks for Semantic Segmentation _FCNs]

[FCN中的transplant][FCN_transplant]

以fcn.berkeleyvision.org的sift-flow-fcn32s为例，训练自己的数据集时除了上面说的要改的地方，还有几个需要注意

1.图像的通道数，如果在VGG上面finetuning的话要求输入通道为3

2.不需要label\_geometric，所以要把与它相关的地方都注释掉

（1）siftflow\_layers.py中搜索label\_geometric，注释掉，if len(top) != 3:改为if len(top) != 2:

（2）trainval.prototxt与test.prototxt中top: "geo"这行注释掉，这个不注释掉会出现check failed:data\_的错误

layer \{  
  name: "loss"  
  type: "SoftmaxWithLoss"  
  bottom: "score\_sem"  
  bottom: "sem"  
  top: "loss"  
  loss\_param \{  
    ignore\_label: 255  
    normalize: false  
  \}  
\}这层以后的和geo相关的层全部注释掉

（3）solve.py中score.seg\_tests(solver, False, test, layer='score\_geo', gt='geo')这句注释掉

3.几个prototxt文件中最后输出的feature map的num\_ouput改成类别数+1

4.改数据输入路径，siftflow\_layers.py中的load\_image和load\_label函数加载图像和label的路径改一下，对应prototxt的param\_str: "\{\\'siftflow\_dir\\':改成自己的数据路径，solve.py中test = np.loadtxt这句改成test.txt的路径

像素级别的loss的取均值决定learning\_rate的数量级. 使用224×224的图片, 如果loss是对单个像素的均值, 则数量级在10−5. 如果是对单张图片的loss, 要再小4个数量级.(caffe fcn使用了10−10)，这里是fcn的[solver.prototxt][]。

## deeplab v2实践 ##

[DeepLab V2 配置bug][DeepLab V2 _bug]

[DeepLab v2 配置][DeepLab v2]

[DeepLab V2安装配置][DeepLab V2]

[crfasrnn环境搭建，跑demo][crfasrnn_demo]

## 资源 ##

https://github.com/shelhamer/fcn.berkeleyvision.org

https://github.com/torrvision/crfasrnn

https://github.com/martinkersner/train-CRF-RNN

https://github.com/wkentaro/fcn

https://github.com/ycszen/pytorch-ss

https://github.com/msracver/FCIS

[Link 1]: http://blog.csdn.net/niaolianjiulin/article/details/52948274
[Link 2]: http://blog.csdn.net/niaolianjiulin/article/details/53098437
[Link 3]: https://mp.weixin.qq.com/s?__biz=MjM5NzY5ODAyNQ==&mid=2651641720&idx=7&sn=8b7c9399b5f140a8f72c67df8e198a83&scene=1&srcid=0915JE56gclgnaiAoWxgrlJr&pass_ticket=OGt4ePT9BBEIomvpJA%2BOWbWToE1DriR1%2Fq8VSweGmHZI1f%2FfVe16utFsOT7Txckb#rd
[Link 4]: https://zhuanlan.zhihu.com/p/22308032
[Link 5]: http://blog.csdn.net/u012938704/article/details/53202103
[FCN]: http://blog.csdn.net/taigw/article/details/51401448
[Link 6]: http://blog.csdn.net/qq_21275321/article/details/77247495
[Fully Convolutional Networks]: http://blog.csdn.net/liyaohhh/article/details/50807423
[Fully Convolutional Networks for Semantic Segmentation]: http://blog.csdn.net/liyaohhh/article/details/50822246
[Fully Convolutional Networks for Semantic Segmentation 1]: http://blog.csdn.net/shenxiaolu1984/article/details/51348149
[FCN_-CVPR 2015]: http://blog.csdn.net/hjimce/article/details/50268555
[FCN 1]: http://www.cnblogs.com/gujianhan/p/6030639.html
[FCN for Sematic Segmentation.pptx]: http://xiahouzuoxin.github.io/notes/enclosure/%E6%B7%B1%E5%BA%A6%E5%8D%B7%E7%A7%AF%E7%BD%91%E7%BB%9CCNN%E4%B8%8E%E5%9B%BE%E5%83%8F%E8%AF%AD%E4%B9%89%E5%88%86%E5%89%B2/FCN%20for%20Sematic%20Segmentation.pptx
[Link 7]: https://zhuanlan.zhihu.com/p/20872103
[Semantic Segmentation with FCN_32x_ 16x_8x]: http://blog.csdn.net/weixin_35653315/article/details/54602190
[--FCN _--]: http://blog.csdn.net/zhangjunhit/article/details/75635476
[deeplab_Hole]: http://blog.csdn.net/tangwei2014/article/details/50453334
[DeepLab]: http://blog.csdn.net/thesby/article/details/52699836
[DeepLab 1]: http://blog.csdn.net/yxq5997/article/details/53693869
[Paper Reading_ DeepLab v1 _ v2]: http://blog.csdn.net/c_row/article/details/52161394
[DeepLab 2]: http://blog.csdn.net/bea_tree/article/details/55096031
[DeepLab_ Semantic Image Segmentation with Deep Convolutional Nets_ Atrous Convolution_ and Fully Connected CRFs]: http://www.dongzhuoyao.com/deeplab-semantic-image-segmentation-with-deep-convolutional-nets-atrous-convolution-and-fully-connected-crfs/
[CRF]: http://blog.csdn.net/thesby/article/details/50969788
[CRF 1]: http://blog.csdn.net/helei001/article/details/40616329
[Conditional Random Fields as Recurrent Neural Networks]: http://blog.csdn.net/meanme/article/details/50838769
[CRF as RNN_-]: http://blog.csdn.net/hjimce/article/details/50888915
[CRF as RNN_Caffe]: http://blog.csdn.net/taigw/article/details/51794283
[FCN_CRF]: http://blog.csdn.net/u012759136/article/details/52434826
[CRF as RNN]: http://blog.csdn.net/thesby/article/details/51593591
[Semantic Image Segmentation With Deep Convolutional Nets and Fully Connected CRFs.ppt]: http://xiahouzuoxin.github.io/notes/enclosure/%E6%B7%B1%E5%BA%A6%E5%8D%B7%E7%A7%AF%E7%BD%91%E7%BB%9CCNN%E4%B8%8E%E5%9B%BE%E5%83%8F%E8%AF%AD%E4%B9%89%E5%88%86%E5%89%B2/Semantic%20Image%20Segmentation%20With%20Deep%20Convolutional%20Nets%20and%20Fully%20Connected%20CRFs.pptx
[-keras]: http://blog.csdn.net/u012931582/article/details/70215756
[Learning Deconvolution Network for Semantic Segmentation]: http://blog.csdn.net/meanme/article/details/50897477
[Instance-aware Semantic Segmentation via Multi-task Network Cascades_MNC]: http://blog.csdn.net/u011070171/article/details/54408303
[IndexError_ index 4 is out of bounds for axis 1 with size 4]: http://blog.csdn.net/u011070171/article/details/70244016
[-_Instance-aware Semantic Segmentation via Multi-task Network Cascades]: http://blog.csdn.net/u011070171/article/details/53071216
[Instance-aware Semantic Segmentation via Multi-task Network Cascades]: http://blog.csdn.net/cv_family_z/article/details/70229640
[ParseNet]: http://blog.csdn.net/thesby/article/details/52701161
[semantic segmentation_ _-FCN SegNet U-net DeconvNet]: http://blog.csdn.net/u012931582/article/details/70314859
[DL_Semantic Segmentation]: http://blog.csdn.net/u011148330/article/details/49446443
[PSPnet]: http://blog.csdn.net/yxq5997/article/details/53695779
[PSPnet_Pyramid Scene Parsing Network]: http://blog.csdn.net/bea_tree/article/details/56678560
[Pyramid Scene Parsing Network - arxiv - 1612.01105]: http://blog.csdn.net/zimenglan_sysu/article/details/53536350
[Link 8]: http://blog.csdn.net/u012759136/article/details/59115476
[CNN]: https://zhuanlan.zhihu.com/p/23811946
[_ BoxSup_ Exploiting Bounding Boxes to Supervise Convolutional Networks for Semantic Segmentati]: http://blog.csdn.net/bea_tree/article/details/52156632
[Link 9]: https://www.zhihu.com/question/53263115/answer/134797599
[Link 10]: https://zhuanlan.zhihu.com/p/25201511
[Learning to Refine Object Segments]: http://blog.csdn.net/u012905422/article/details/52353361
[Mutil-Scale CNN_-ICCV 2015]: http://blog.csdn.net/hjimce/article/details/50443995
[Paper Reading_ Attention to Scale_ Scale-aware Semantic Image Segmentation]: http://blog.csdn.net/c_row/article/details/52092924
[Zifeng Wu_38_Wider or Deeper_ Revisiting the ResNet Model for Visual Recognition]: http://blog.csdn.net/bea_tree/article/details/56282665
[RefineNet_RefineNet_ Multi-Path Refinement Networks for High-Resolution Semantic Segmentation]: http://blog.csdn.net/bea_tree/article/details/58208386
[zifeng wu instance level segmentation_Bridging Category-level and Instance-level Semantic Image]: http://blog.csdn.net/bea_tree/article/details/59480787
[-Large Kernel Matters _ Improve Semantic Segmentation by Global Convolutional Network]: http://blog.csdn.net/bea_tree/article/details/60977512
[Label Refinement Network for Coarse-to-Fine Semantic Segmentation]: http://blog.csdn.net/u012905422/article/details/61614728
[Fully Convolutional Instance-aware Semantic Segmentation]: http://coolwyj.com/fully-convolutional-instance-aware-semantic-segmentation-%E8%AE%BA%E6%96%87%E7%AC%94%E8%AE%B0/
[- Semantic Segmentation Papers]: http://blog.csdn.net/zziahgf/article/details/72639791
[Ubuntu_FCN]: http://www.itdadao.com/articles/c15a530243p0.html
[FCN_]: http://blog.csdn.net/u013059662/article/details/52770198
[FCN 2]: http://www.itdadao.com/articles/c15a229744p0.html
[net_surgery]: http://blog.csdn.net/zy3381/article/details/50470382
[caffe_FCN_fine-tune]: http://blog.csdn.net/zy3381/article/details/50458331
[FCN_SIFT-Flow]: http://www.cnblogs.com/xuanxufeng/p/6243342.html
[FCN 3]: http://blog.csdn.net/wangkun1340378/article/details/56834642
[FCN_ SIFTFLOW]: http://blog.csdn.net/wangkun1340378/article/details/70238290
[Tensoflow_13--_Fully Convolutional Networks for Semantic Segmentation _FCNs]: http://blog.csdn.net/best_coder/article/details/70146441
[FCN_transplant]: http://blog.csdn.net/iamzhangzhuping/article/details/71566439
[solver.prototxt]: https://github.com/shelhamer/fcn.berkeleyvision.org/blob/master/voc-fcn32s/solver.prototxt#L10
[DeepLab V2 _bug]: http://blog.csdn.net/liuxiabing150/article/details/52035621
[DeepLab v2]: http://blog.csdn.net/xczexcel/article/details/70185643
[DeepLab V2]: http://blog.csdn.net/tianrolin/article/details/71246472
[crfasrnn_demo]: http://blog.csdn.net/liuxiabing150/article/details/51636650