The Star Also Rises: November 2019

Monday, November 25, 2019

Semantic Segmentation

Semantic Segmentation

2019/01/17

-----

Fig. 1. FCN（對每個像素進行分類） [1]。

-----

# ICNet

-----

References

Paper

# ICNet

Zhao, Hengshuang, et al. "Icnet for real-time semantic segmentation on high-resolution images." Proceedings of the European Conference on Computer Vision (ECCV). 2018.

http://openaccess.thecvf.com/content_ECCV_2018/papers/Hengshuang_Zhao_ICNet_for_Real-Time_ECCV_2018_paper.pdf

# Survey

Garcia-Garcia, Alberto, et al. "A survey on deep learning techniques for image and video semantic segmentation." Applied Soft Computing 70 (2018): 41-65.

https://www.sciencedirect.com/science/article/pii/S1568494618302813

-----

A 2019 Guide to Semantic Segmentation - Heartbeat
https://heartbeat.fritz.ai/a-2019-guide-to-semantic-segmentation-ca8242f5a7fc

Semantic Segmentation _ Zhang Bin's Blog
https://zhangbin0917.github.io/2018/09/18/Semantic-Segmentation/

GitHub - mrgloom_awesome-semantic-segmentation awesome-semantic-segmentation
https://github.com/mrgloom/awesome-semantic-segmentation

GitHub - GeorgeSeif_Semantic-Segmentation-Suite Semantic Segmentation Suite in TensorFlow. Implement, train, and test new Semantic Segmentation models easily!
https://github.com/GeorgeSeif/Semantic-Segmentation-Suite

An overview of semantic image segmentation
https://www.jeremyjordan.me/semantic-segmentation/

Semantic Segmentation with Deep Learning – Towards Data Science
https://towardsdatascience.com/semantic-segmentation-with-deep-learning-a-guide-and-code-e52fc8958823

Semantic Segmentation using Fully Convolutional Networks over the years
https://meetshah1995.github.io/semantic-segmentation/deep-learning/pytorch/visdom/2017/06/01/semantic-segmentation-over-the-years.html

-----

A 2017 Guide to Semantic Segmentation with Deep Learning
http://blog.qure.ai/notes/semantic-segmentation-deep-learning-review

速記AI課程－Convolutional Neural Networks for Computer Vision Applications（二）
https://medium.com/@baubibi/%E9%80%9F%E8%A8%98ai%E8%AA%B2%E7%A8%8B-convolutional-neural-networks-for-computer-vision-applications-%E4%BA%8C-d5fbb995ffd7

从全卷积网络到大型卷积核：深度学习的语义分割全指南 _ 机器之心
https://www.jiqizhixin.com/articles/2017-07-14-10

Going beyond the bounding box with semantic segmentation
https://www.jiqizhixin.com/articles/2018-06-04-17

一文了解什么是语义分割及常用的语义分割方法有哪些 _ 机器之心
https://www.jiqizhixin.com/articles/2018-06-04-17

深度学习（十九）——FCN, SegNet, DeconvNet, DeepLab, ENet, GCN - antkillerfarm的专栏 - CSDN博客
https://blog.csdn.net/antkillerfarm/article/details/79524417

语义分割 _ 发展综述 - 知乎
https://zhuanlan.zhihu.com/p/37618829

BlitzNet

2019/11/18

前言：

有別於 RefineNet 與 PSPNet，BlitzNet 採取 DSSD 的架構，著重於速度。但 RefineNet 分別有 RefineNet-LW 輕量化與 RefineNet-AA 輕量化加上景深預測，而 PSPNet 之後則有 ICNet、BiSeNet 與 Fast-SCNN 的快速模型提出，因此 BlitzNet 的重要性已減低。

-----

# BlitzNet

-----

# DSSD

-----

# BlitzNet

-----

# BlitzNet

-----

References

# BlitzNet

Dvornik, Nikita, et al. "Blitznet: A real-time deep network for scene understanding." Proceedings of the IEEE international conference on computer vision. 2017.

http://openaccess.thecvf.com/content_ICCV_2017/papers/Dvornik_BlitzNet_A_Real-Time_ICCV_2017_paper.pdf

# DSSD

Fu, Cheng-Yang, et al. "Dssd: Deconvolutional single shot detector." arXiv preprint arXiv:1701.06659 (2017).

https://arxiv.org/pdf/1701.06659.pdf

-----

BlitzNet A Real-Time Deep Network for Scene Understanding - ALISURE
https://alisure.github.io/2018/03/27/Paper/Semantic-Segmentation-BlitzNet/

Fast-SCNN

Fast-SCNN

2019/11/21

前言：

Fast-SCNN，通過融合 two-branch 方法和經典的 encoder-decoder 方法，達到了 real-time 的效果。深度卷積網路中的前幾個層提取的是低階段特徵，藉鑑了 two-branch 的方法，將前幾個層的計算進行共享稱為「學習下採樣」（learn to downsample），作用類似 encoder-decoder 中的 identity mapping。最後再加上高效的 depthwise separable 卷積和 inverse Residual blocks。

-----

# Fast-SCNN

-----

# Fast-SCNN

-----

References

# Fast-SCNN

Poudel, Rudra PK, Stephan Liwicki, and Roberto Cipolla. "Fast-SCNN: fast semantic segmentation network." arXiv preprint arXiv:1902.04502 (2019).

https://arxiv.org/pdf/1902.04502.pdf

Fast-SCNN Fast Semantic Segmentation Network 论文学习 - calvinpaean的博客 - CSDN博客
https://blog.csdn.net/calvinpaean/article/details/88534052

BiSeNet

BiSeNet

2019/11/25

前言：

BiSeNet 沿襲了 PSPNet 的空間金字塔與 ICNet 的 multi-branch，提出了包含 Spatial 與 Context 的 two-branch 架構，並且用特徵融合模塊 FFM 整合這兩路的特徵以及 BiSeNet 的整體架構。

-----

# BiSeNet

-----

# BiSeNet

-----

-----

References

# BiSeNet

Yu, Changqian, et al. "Bisenet: Bilateral segmentation network for real-time semantic segmentation." Proceedings of the European Conference on Computer Vision (ECCV). 2018.

http://openaccess.thecvf.com/content_ECCV_2018/papers/Changqian_Yu_BiSeNet_Bilateral_Segmentation_ECCV_2018_paper.pdf

旷视科技提出双向网络BiSeNet：实现实时语义分割 - 知乎
https://zhuanlan.zhihu.com/p/41475332

ICNet

ICNet

2019/11/21

前言：

ICNet 的作者首先回顧了 PSPNet，然後引入了加速語義分割的直觀策略。先讓低分辨率的圖像先通過網路得到粗略的預測圖，然後在 cascade fusion unit 導入中等分辨率和高分辨率的圖像逐漸改善預測圖。採用 multi-branch 的架構，在速度與正確性上，取得了不錯的平衡。

-----

# ICNet

-----

# ICNet

-----

References

# ICNet

Zhao, Hengshuang, et al. "Icnet for real-time semantic segmentation on high-resolution images." Proceedings of the European Conference on Computer Vision (ECCV). 2018.

http://openaccess.thecvf.com/content_ECCV_2018/papers/Hengshuang_Zhao_ICNet_for_Real-Time_ECCV_2018_paper.pdf

论文阅读 - ICNet for Real-Time Semantic Segmentation on High-Resolution Images(br)(ECCV 2018 CUHK, SenseTime) _ Zhang Bin's Blog
https://zhangbin0917.github.io/2018/05/29/ICNet-for-Real-Time-Semantic-Segmentation-on-High-Resolution-Images/

PSPNet

PSPNet

2019/10/14

前言：

PSPNet 主要在 ResNet 上使用 FPN 的金字塔，多尺度特徵圖，這個概念，另外加上輔助的損失函數，完成對細部物件的語意分割。

-----

# PSPNet

-----

# FPN

-----

# PSPNet

-----

# PSPNet

-----

References

# PSPNet

Zhao, Hengshuang, et al. "Pyramid scene parsing network." Proceedings of the IEEE conference on computer vision and pattern recognition. 2017.

http://openaccess.thecvf.com/content_cvpr_2017/papers/Zhao_Pyramid_Scene_Parsing_CVPR_2017_paper.pdf

# FPN

Lin, Tsung-Yi, et al. "Feature pyramid networks for object detection." CVPR. Vol. 1. No. 2. 2017.

http://openaccess.thecvf.com/content_cvpr_2017/papers/Lin_Feature_Pyramid_Networks_CVPR_2017_paper.pdf

-----

Review PSPNet — Winner in ILSVRC 2016 (Semantic Segmentation _ Scene Parsing)
https://towardsdatascience.com/review-pspnet-winner-in-ilsvrc-2016-semantic-segmentation-scene-parsing-e089e5df177d

Pyramid Scene Parsing Network - 知乎
https://zhuanlan.zhihu.com/p/36670958

PSPNet ——语义分割及场景分析 - 云+社区 - 腾讯云
https://cloud.tencent.com/developer/article/1491359

RefineNet-AA

RefineNet-AA

2019/11/21

前言：

RefineNet-AA 在 Light-weight RefineNet（RefineNet-LW）的基礎上，以 asymmetric annotations 的方法，增加了預測影像深度的功能。

-----

# RefineNet-AA

-----

# RefineNet-AA

-----

References

# RefineNet-AA

Nekrasov, Vladimir, et al. "Real-time joint semantic segmentation and depth estimation using asymmetric annotations." 2019 International Conference on Robotics and Automation (ICRA). IEEE, 2019.

https://arxiv.org/pdf/1809.04766.pdf

17毫秒每帧！实时语义分割与深度估计
https://mp.weixin.qq.com/s/Sn3N8IxHtgp53Y0VLIPJCQ?utm_source=tuicool&utm_medium=referral

RefineNet-LW

RefineNet-LW

2019/11/21

前言：

Light-weight RefineNet 藉由三個改進讓 RefineNet 適合在行動裝置上運作。首先是將 Conv3 改成 Conv1，其次是去除殘差卷積單元，最後則是將網路骨幹改成行動裝置的卷積網路如 MobileNet 等。

-----

# RefineNet-LW

-----

# RefineNet-LW

-----

References

# RefineNet-LW

Nekrasov, Vladimir, Chunhua Shen, and Ian Reid. "Light-weight refinenet for real-time semantic segmentation." arXiv preprint arXiv:1810.03272 (2018).

https://arxiv.org/pdf/1810.03272.pdf

图像分割之Light-Weight RefineNet - 年轻即出发， - CSDN博客
https://blog.csdn.net/qq_14845119/article/details/84776977

RefineNet

RefineNet

2019/11/18

前言：

RefineNet 可以參考 RefineDet 的兩路設計。首先第一路的卷積網路可以產生不同大小的特徵圖，再進入第二路的 Refine 模組運算。模組主要由殘差卷積單元、多解析融合、鍊式殘差池化、以及輸出層構成。

-----

# RefineNet

-----

# RefineDet

-----

# RefineNet

-----

// Review RefineNet — Multi-path Refinement Network (Semantic Segmentation)

-----

# RefineNet

-----

References

# RefineNet

Lin, Guosheng, et al. "Refinenet: Multi-path refinement networks for high-resolution semantic segmentation." Proceedings of the IEEE conference on computer vision and pattern recognition. 2017.

http://openaccess.thecvf.com/content_cvpr_2017/papers/Lin_RefineNet_Multi-Path_Refinement_CVPR_2017_paper.pdf

# RefineDet

Zhang, Shifeng, et al. "Single-shot refinement neural network for object detection." Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2018.

http://openaccess.thecvf.com/content_cvpr_2018/papers/Zhang_Single-Shot_Refinement_Neural_CVPR_2018_paper.pdf

-----

Review RefineNet — Multi-path Refinement Network (Semantic Segmentation)
https://towardsdatascience.com/review-refinenet-multi-path-refinement-network-semantic-segmentation-5763d9da47c1

-----

语义分割之RefineNet - 知乎
https://zhuanlan.zhihu.com/p/37805109

ResNet-38

ResNet-38

2019/11/18

前言：

ResNet-38 由論文作者調整 ResNet 的寬度與深度，打造一個適合語意分割的卷積網路。

-----

# ResNet-38

-----

# ResNet-38

-----

References

# ResNet-38

Wu, Zifeng, Chunhua Shen, and Anton Van Den Hengel. "Wider or deeper: Revisiting the resnet model for visual recognition." Pattern Recognition 90 (2019): 119-133.

https://arxiv.org/pdf/1611.10080.pdf

segmentation论文总结：Deeplab-v2; FCN-xs; Resnet based FCN - 简书
https://www.jianshu.com/p/b09ae7195790