CRF learning with CNN features for image segmentation

作者：

Highlights：

• A deep CNN pretrained on ImageNet generalizes well to various segmentation datasets.

• Deep features significantly outperform BoW and unsupervisd feature learning.

• Combining deep CNN features with CRF yields new state-of-the-art results.

• Incorporating spatial related co-occurrence potentials further improves the accuracy.

摘要

Highlights•A deep CNN pretrained on ImageNet generalizes well to various segmentation datasets.•Deep features significantly outperform BoW and unsupervisd feature learning.•Combining deep CNN features with CRF yields new state-of-the-art results.•Incorporating spatial related co-occurrence potentials further improves the accuracy.

论文关键词：Conditional random field (CRF),Convolutional neural network (CNN),Structured support vector machine (SSVM),Co-occurrence

论文评审过程：Received 9 September 2014, Revised 28 March 2015, Accepted 17 April 2015, Available online 24 April 2015, Version of Record 17 June 2015.

论文官网地址：https://doi.org/10.1016/j.patcog.2015.04.019