READING NOTE: Learning Deconvolution Network for Semantic Segmentation
2016-07-04 22:57
531 查看
TITLE: Learning Deconvolution Network for Semantic Segmentation
AUTHER: Hyeonwoo Noh, Seunghoon Hong, Bohyung Han
ASSOCIATION: Department of Computer Science and Engineering, POSTECH, Korea
FROM: arXiv:1505.04366
Instance-wise segmentations are merged for final sematic segmentation, which is free from scale issues.
Object proposals are genereated by alogrithms such as EdgeBox.
ROI extracted based on object proposals are sent to the Deconvolution Network. The outputs are instance-wise segmentations.
instance-wise segmentations are combined to get the final segmentaton.
Some Details
Architecture of the network is shown as the following figure. In the network, unpooling operation captures example-specific structures by tracing the original locations with strong activations back to image space. On the other hand, deconvolution operation learnes filters to capture class-specific shapes.
Training contains two stages. At first stage, simpler data are used to train the network. The simpler data are generated using object annotations and contains constraint appearance of objects. At second stage, complex data are similarly generated but from object proposals.
Inference includes a CRF can further bootstrap the performance.
Deconvolution can generate finer segmentations.
AUTHER: Hyeonwoo Noh, Seunghoon Hong, Bohyung Han
ASSOCIATION: Department of Computer Science and Engineering, POSTECH, Korea
FROM: arXiv:1505.04366
CONTRIBUTIONS
A multi-layer deconvolution network is designed and learned, which is composed of deconvolution, unpooling, and rectified linear unit (ReLU) layers.Instance-wise segmentations are merged for final sematic segmentation, which is free from scale issues.
METHOD
The main steps of the method is as follows:Object proposals are genereated by alogrithms such as EdgeBox.
ROI extracted based on object proposals are sent to the Deconvolution Network. The outputs are instance-wise segmentations.
instance-wise segmentations are combined to get the final segmentaton.
Some Details
Architecture of the network is shown as the following figure. In the network, unpooling operation captures example-specific structures by tracing the original locations with strong activations back to image space. On the other hand, deconvolution operation learnes filters to capture class-specific shapes.
Training contains two stages. At first stage, simpler data are used to train the network. The simpler data are generated using object annotations and contains constraint appearance of objects. At second stage, complex data are similarly generated but from object proposals.
Inference includes a CRF can further bootstrap the performance.
ADVANTAGES
It handles objects in various scales effectively and identifies fine details of objects .Deconvolution can generate finer segmentations.
DISADVANTAGES
Large number of proposals are needed to get better result, which means higher computational complexity.相关文章推荐
- leetcode: Candy
- 牛逼的OSQL----大数据导入
- centos7 挂载数据盘
- SSI
- Livable City
- JavaScript学习总结(五)原型和原型链详解
- Java CopyOnWriteArrayList详解
- net常用命令
- 在线生成ICO图标
- Excel工作表保护密码破解方法
- 【记录】寄存器地址映射
- 删除鼠标右键多余的菜单
- windows自带的特殊符号
- 找回丢失的“显示桌面”快速启动按钮
- 用友打印问题合集 一【各模块】
- ZF_20160704_CF15X
- 英语钻石法则(五)-----听说并进:讲故事
- MSDB置疑修复
- 【转】U盘量产解释
- 设置windows服务依存关系