Pyramid Scene Parsing Network

CVPR 2017 Hengshuang ZhaoJianping ShiXiaojuan QiXiaogang WangJiaya Jia

Scene parsing is challenging for unrestricted open vocabulary and diverse scenes. In this paper, we exploit the capability of global context information by different-region-based context aggregation through our pyramid pooling module together with the proposed pyramid scene parsing network (PSPNet)... (read more)

PDF Abstract

Evaluation results from the paper


Task Dataset Model Metric name Metric value Global rank Compare
Semantic Segmentation ADE20K PSPNet Validation mIoU 44.94 # 1
Semantic Segmentation ADE20K PSPNet Test Score 0.5538 # 3
Semantic Segmentation CamVid PSPNet Mean IoU 69.1% # 1
Real-Time Semantic Segmentation CamVid PSPNet mIoU 69.1% # 1
Real-Time Semantic Segmentation CamVid PSPNet Time (ms) 185 # 2
Real-Time Semantic Segmentation CamVid PSPNet Frame (fps) 5.4 # 2
Semantic Segmentation Cityscapes PSPNet Mean IoU 81.2% # 5
Real-Time Semantic Segmentation Cityscapes PSPNet mIoU 81.2% # 1
Real-Time Semantic Segmentation Cityscapes PSPNet Time (ms) 1288 # 7
Real-Time Semantic Segmentation Cityscapes PSPNet Frame (fps) 0.78 # 9
Semantic Segmentation PASCAL VOC 2012 PSPNet Mean IoU 85.4% # 5