Reference Twice: A Simple and Unified Baseline for Few-Shot Instance Segmentation

TL;DR

This paper introduces Reference Twice, a transformer-based unified framework for Few-Shot Instance Segmentation that effectively leverages support-query relationships, avoids overfitting, and extends to incremental learning with improved performance.

Contribution

The paper presents a novel transformer-based baseline for FSIS, simplifies support-query interaction, and proposes a class-enhanced distillation loss for incremental learning.

Findings

Achieves +8.2/+9.4 performance gain over state-of-the-art on COCO.

Effectively avoids overfitting in FSIS.

Enables easy extension to incremental FSIS.

Abstract

Few-Shot Instance Segmentation (FSIS) requires detecting and segmenting novel classes with limited support examples. Existing methods based on Region Proposal Networks (RPNs) face two issues: 1) Overfitting suppresses novel class objects; 2) Dual-branch models require complex spatial correlation strategies to prevent spatial information loss when generating class prototypes. We introduce a unified framework, Reference Twice (RefT), to exploit the relationship between support and query features for FSIS and related tasks. Our three main contributions are: 1) A novel transformer-based baseline that avoids overfitting, offering a new direction for FSIS; 2) Demonstrating that support object queries encode key factors after base training, allowing query features to be enhanced twice at both feature and query levels using simple cross-attention, thus avoiding complex spatial correlation interaction; 3) Introducing a class-enhanced base knowledge distillation loss to address the issue of DETR-like models struggling with incremental settings due to the input projection layer, enabling easy extension to incremental FSIS. Extensive experimental evaluations on the COCO dataset under three FSIS settings demonstrate that our method performs favorably against existing approaches across different shots, \eg, $+8.2/+9.4$ performance gain over state-of-the-art methods with 10/30-shots. Source code and models will be available at https://github.com/hanyue1648/RefT.