on arxiv: https://arxiv.org/abs/1806.07373

by Kate Rakelly*, Evan Shelhamer*, Trevor Darrell, Alexei A. Efros, and Sergey Levine UC Berkeley

Learning-based methods for visual segmentation have made progress on particular types of segmentation tasks, but are limited by the necessary supervision, the narrow definitions of fixed tasks, and the lack of control during inference for correcting errors. To remedy the rigidity and annotation burden of standard approaches, we address the problem of few-shot segmentation: given few image and few pixel supervision, segment any images accordingly. We propose guided networks, which extract a latent task representation from any amount of supervision, and optimize our architecture end-to-end for fast, accurate few-shot segmentation. Our method can switch tasks without further optimization and quickly update when given more guidance. We report the first results for segmentation from one pixel per concept and show real-time interactive video segmentation. Our unified approach propagates pixel annotations across space for interactive segmentation, across time for video segmentation, and across scenes for semantic segmentation. Our guided segmentor is state-of-the-art in accuracy for the amount of annotation and time.

This is a work-in-progress, not yet a reference implementation of the paper, and could change at any time.

• for few-shot interactive image segmentation and few-shot semantic segmentation, see this branch.
• for few-shot video object segmentation, see branch video-seg (note: this is older code, for pytorch 0.3.1)

TODO

• port to pytorch 1.0
• push branch for interactive video segmentation
• reconcile branches into unified implementation of few-shot segmentation

Please check back soon for improvements, pre-trained models, and usage notebooks!