「Computer Vision」Note on Decoders Matter for Semantic Segmentation

版权声明：本文为博主原创文章，未经博主允许不得转载。 https://blog.csdn.net/dgyuanshaofeng/article/details/88320903

QQ Group: 428014259
Tencent E-mail：[email protected]
http://blog.csdn.net/dgyuanshaofeng/article/details/88320903

作者：Zhi Tian, Tong He, Chunhua Shen, Youliang Yan
单位：The University of Adelaide, Australia；Noah’s Ark Lab, Huawei Technologies

0 摘要

指出双线性上采样过于简单，数据不依赖。
提出数据依赖上采样取代双线性上采样，其利用的是标签空间的冗余性。
性能： 88.1% mIOU on PASCAL VOC with 30% computation of the previously best model；52.5% mIOU on PASCAL Context

1 介绍

[1] Decoders Matter for Semantic Segmentation: Data-Dependent Decoding Enables Flexible Feature Aggregation CVPR 2019 [paper]