DeepLab v3
2019/11/18
前言:
DeepLab v3:採用多比例的帶孔卷積級聯或併行來捕獲多尺度背景,基於圖像特徵優化 ASPP。並取消在 v1 與 v2 所使用的 FC-CRF。
-----
# DeepLab v3
-----
# DeepLab v3
-----
# DeepLab v3
-----
# DeepLab v3
-----
References
# DeepLab v3
Chen, Liang-Chieh, et al. "Rethinking atrous convolution for semantic image segmentation." arXiv preprint arXiv:1706.05587 (2017).
https://arxiv.org/pdf/1706.05587.pdf DeepLab V3 论文笔记 - 知乎
https://zhuanlan.zhihu.com/p/40470298
No comments:
Post a Comment