SAIS: Single-stage Anchor-free Instance Segmentation

TL;DR

SAIS introduces a simple, efficient single-stage anchor-free method for instance segmentation that combines mask prototypes and coefficients, utilizing fused information and center-aware targets to improve accuracy with less memory.

Contribution

The paper presents a novel single-stage anchor-free instance segmentation approach that enhances mask quality through fused features and center-aware targets, achieving state-of-the-art performance with reduced memory.

Findings

Achieves competitive performance on MS COCO

Uses less memory than existing methods

Improves mask quality with fused features

Abstract

In this paper, we propose a simple yet efficientinstance segmentation approach based on the single-stage anchor-free detector, termed SAIS. In our approach, the instancesegmentation task consists of two parallel subtasks which re-spectively predict the mask coefficients and the mask prototypes.Then, instance masks are generated by linearly combining theprototypes with the mask coefficients. To enhance the quality ofinstance mask, the information from regression and classificationis fused to predict the mask coefficients. In addition, center-aware target is designed to preserve the center coordination ofeach instance, which achieves a stable improvement in instancesegmentation. The experiment on MS COCO shows that SAISachieves the performance of the exiting state-of-the-art single-stage methods with a much less memory footpr