Monday, November 25, 2019

DeepLab v2

DeepLab v2

2019/11/18

前言:

DeepLab v2:基於 DeepLab v1 的優化,使用空洞空間金字塔池化(atrous spatial pyramid pooling,ASPP),再以 FC-CRF 最佳化,對物體進行有效的分割 。

-----


# DeepLab v2

-----


# DeepLab v2

-----


# DeepLab v2

-----


# DeepLab v2

-----


# DeepLab v2

-----

References

# DeepLab v2
Chen, Liang-Chieh, et al. "Deeplab: Semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected crfs." arXiv preprint arXiv:1606.00915 (2016).
https://arxiv.org/pdf/1606.00915.pdf 

DeepLab V2 论文笔记 - 知乎
https://zhuanlan.zhihu.com/p/37645755

No comments: