One-Shot Learning for Semantic Segmentation

About

Low-shot learning methods for image classification support learning from sparse data. We extend these techniques to support dense semantic image segmentation. Specifically, we train a network that, given a small set of annotated images, produces parameters for a Fully Convolutional Network (FCN). We use this FCN to perform dense pixel-level prediction on a test image for the new semantic class. Our architecture shows a 25% relative meanIoU improvement compared to the best baseline methods for one-shot segmentation on unseen classes in the PASCAL VOC 2012 dataset and is at least 3 times faster.

Amirreza Shaban, Shray Bansal, Zhen Liu, Irfan Essa, Byron Boots · 2017

Related benchmarks

Task	Dataset	Result
Few-shot Segmentation	PASCAL-5i	mIoU (Fold 0)35.9	325
Semantic segmentation	COCO 2014 (val)	mIoU46.2	304
Few-shot Semantic Segmentation	PASCAL-5^i (test)	FB-IoU61.5	177
Semantic segmentation	PASCAL-5i	Mean mIoU43.9	149
Semantic segmentation	PASCAL-5^i (test)	Mean Score44	107
Semantic segmentation	PASCAL 5-shot 5i	Mean mIoU43.95	100
Few-shot Segmentation	PASCAL 5i (val)	mIoU (Mean)40.8	83
Semantic segmentation	PASCAL-5^i Fold-1	mIoU58.1	75
Semantic segmentation	PASCAL-5^i Fold-3	mIoU39.1	75
Semantic segmentation	PASCAL-5^i Fold-2	mIoU42.7	75