专注语义分割系列技术

1) Deeplab V3和Deeplab V3+

      https://blog.csdn.net/fish_like_apple/article/details/82787705

2) Pyramid Scene Parsing Net

       https://blog.csdn.net/u011974639/article/details/78985130

3) ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile Devices

        https://blog.csdn.net/c602273091/article/details/75137878

4）Xception: Deep learning with depthwise separable convolutions

    https://blog.csdn.net/albertshi12/article/details/81303245

5) deformable convolutional networks – coco detection and segmentation challenge 2017 entry

        https://blog.csdn.net/qq_21949357/article/details/80538255

6）非论文，Depthwise Separable Convolution详解，非常清晰

    https://blog.csdn.net/sinat_26114733/article/details/89076714

7) BinaryNet

https://blog.csdn.net/jackytintin/article/details/53445280

https://www.cnblogs.com/huxiaozhouzhou/p/10632510.html

8) MobileNet

https://blog.csdn.net/mzpmzk/article/details/82976871

9) Fully Convolutional Instance-aware Semantic Segmentation

http://openaccess.thecvf.com/content_cvpr_2017/papers/Li_Fully_Convolutional_Instance-Aware_CVPR_2017_paper.pdf

(10) FCN:  Fully Convolutional Networks for semantic Segmentation

     中文翻译： https://www.cnblogs.com/xuanxufeng/p/6249834.html

spd U策略：1）使用空洞卷积;2）使用depthwise  convolution