Fully Convolutional Networks for Semantic Segmentation

CVPR 2015 Evan ShelhamerJonathan LongTrevor Darrell

Convolutional networks are powerful visual models that yield hierarchies of features. We show that convolutional networks by themselves, trained end-to-end, pixels-to-pixels, improve on the previous best result in semantic segmentation... (read more)

PDF Abstract

Evaluation results from the paper


Task Dataset Model Metric name Metric value Global rank Compare
Real-Time Semantic Segmentation Cityscapes FCN mIoU 65.3% # 7
Real-Time Semantic Segmentation Cityscapes FCN Time (ms) 500 # 5
Real-Time Semantic Segmentation Cityscapes FCN Frame (fps) 2 # 7
Semantic Segmentation Cityscapes FCN Mean IoU 65.3% # 14
Semantic Segmentation PASCAL VOC 2012 FCN Mean IoU 67.2% # 18
Scene Segmentation SUN-RGBD FCN Mean IoU 27.39 # 3