[zoom-out]Feedforward semantic segmentation with zoom-out features
2018-03-06 20:02
453 查看
Abstract
CVPR 2015的文章,作者来自芝加哥丰田技术学院。早期的方法大多基于随机场结构来获取结构信息,本文考虑不使用这些方法来解决分割问题. 本文的语义分割方法是基于超像素级别的,主要做法就是使用缩放结构来利用不同等级的空间特征对超像素的类别进行判定,从而达到分割的目的.Framework
local zoom
即所求超像素块
proximal zoom
超像素块为中心半径为2的区域
distant zoom
更大的一个bounding box
global zoom
整张图片
本文的主要做法就是对于local和proximal使用手工提取特征,对于distant和global使用conv提取特征,最终形成一个特征如下:
然后对这个总的特征训练一个分类器,损失函数经过了平衡,如下:
通过对每一个超像素块进行分类来实现分割
Result
Others
没有开源相关文章推荐
- 论文笔记《Feedforward semantic segmentation with zoom-out features》
- DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution--阅读笔记
- A 2017 Guide to Semantic Segmentation with Deep Learning
- A 2017 Guide to Semantic Segmentation with Deep Learning 笔记
- 【deeplab】Semantic Image Segmentation with Deep Convolutional Nets and Fully
- 深度学习论文(八)---DeepLabV1-SEMANTIC IMAGE SEGMENTATION WITH DEEP CONVOLUTIONAL NETS AND FULLY CONNECTED C
- DeepLab:深度卷积网络,多孔卷积 和全连接条件随机场 的图像语义分割 Semantic Image Segmentation with Deep Convolutional Nets, Atro
- Semantic Segmentation with FCN里的32x, 16x和8x
- 《Semantic image segmentation with deep convolution nets and fully connected CRFs》学习总结(deeplabV1)
- 语义分割DeepLab v2--DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolut
- 论文阅读-《Semantic Segmentation with Reverse Attention》
- SPD matrix features: Semantic segmentation with second-order pooling
- Semantic Segmentation -- (DeepLabv1)Semantic image segmentation with deep convolutional ... CRFs论文解读
- 论文阅读理解 - Semantic Image Segmentation With Deep Convolutional Nets and Fully Connected CRFs
- 深度学习论文(九)---DeepLabV2-Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution,
- Classification with an edge: Improving semantic image segmentation with boundary detection
- Rich feature hierarchies for accurate object detection and semantic segmentation(译)
- [2016,CVPR] Learning Deep Feature Representations with Domain Guided Dropout for Person Re-id
- [Paper note] RefineNet: Multi-Path Refinement Networks for High-Resolution Semantic Segmentation
- Configuring LabVIEW, LabVIEW DSC, NI Variable Engine, and Lookout to Work with the Windows Firewall