![](https://img-blog.csdnimg.cn/20201014180756928.png?x-oss-process=image/resize,m_fixed,h_64,w_64)
语义分割
老光头_ME2CS
机械转计算机视觉的光头老僧
展开
-
【语义分割】DANet Dual Attention Network for Scene Segmentation
A∈RC×H×WA\in \mathbb{R}^{C \times H \times W}A∈RC×H×W{B,C}∈RC×H×W\{\mathbf{B}, \mathbf{C}\} \in \mathbb{R}^{C \times H \times W}{B,C}∈RC×H×WRC×N\mathbb{R}^{C \times N}RC×Nsji=exp(Bi⋅Cj)∑i=1Nexp(Bi⋅Cj)s_{j i}=\frac{\exp \left(B_{i} \cdot C_{j}\right)}{原创 2020-10-07 20:19:35 · 1408 阅读 · 1 评论 -
【语义分割】 DRANet Scene Segmentation With Dual Relation-Aware Attention Network
2019 CVPR DANet (Dual Attention Network for Scene Segmentation)进阶版,或者轻量化计算量版本。原文中表述DANet的问题是虽然没有增加模型参数,但是每个点间和通道间的相关系数求解增加了模型的计算量和GPU内存使用attention modeling brings a heavy burdern on computation and memory if the number of pixels/channels is huge于是乎,由原原创 2020-10-31 11:52:00 · 3064 阅读 · 6 评论