原文题目:Learning Deconvolution Network for Semantic Segmentation( ICCV, 2015

开源地址:http://cvlab.postech.ac.kr/research/deconvnet/

本篇文章主要是做图像分割,和Fully Convolutional Networks for Semantic Segmentation有些类似。

不同之点在于,本文构造了一个Convolution-deconvolution网络,convolution层主要用于特征提取和泛化,而deconvolution层,添加了特殊的unpooling层,相当于pooling的逆操作,还原图像大小。

整体网络结构如下:

Learning Deconvolution Network for Semantic Segmentation

Learning Deconvolution Network for Semantic SegmentationLearning Deconvolution Network for Semantic Segmentation

结果如下Learning Deconvolution Network for Semantic Segmentation
Learning Deconvolution Network for Semantic Segmentation

相关文章: