Make It Up: Fake Images, Real Gains in Generalized Few-shot Semantic Segmentation

2026-03-28

AI Summary

This work addresses the challenges of generalized few-shot semantic segmentation, where limited annotations for novel classes lead to insufficient appearance coverage and noisy pseudo-labels. To tackle these issues, the authors propose Syn4Seg, a framework that leverages a deduplicated prompt library to generate diverse yet class-consistent synthetic images. It further introduces a support-set-guided two-stage pseudo-label refinement mechanism and a boundary-constrained SAM updating strategy. By integrating diffusion models, adaptive local-global prototype learning, and high-quality synthetic data, Syn4Seg achieves state-of-the-art performance under both 1-shot and 5-shot settings on the PASCAL-5^i and COCO-20^i benchmarks, significantly improving coverage of novel categories and segmentation boundary accuracy.

Abstract

Generalized few-shot semantic segmentation (GFSS) is fundamentally limited by the coverage of novel-class appearances under scarce annotations. While diffusion models can synthesize novel-class images at scale, practical gains are often hindered by insufficient coverage and noisy supervision when masks are unavailable or unreliable. We propose Syn4Seg, a generation-enhanced GFSS framework designed to expand novel-class coverage while improving pseudo-label quality. Syn4Seg first maximizes prompt-space coverage by constructing an embedding-deduplicated prompt bank for each novel class, yielding diverse yet class-consistent synthetic images. It then performs support-guided pseudo-label estimation via a two-stage refinement that i) filters low-consistency regions to obtain high-precision seeds and ii) relabels uncertain pixels with image-adaptive prototypes that combine global (support) and local (image) statistics. Finally, we refine only boundary-band and unlabeled pixels using a constrained SAM-based update to improve contour fidelity without overwriting high-confidence interiors. Extensive experiments on PASCAL-$5^i$ and COCO-$20^i$ demonstrate consistent improvements in both 1-shot and 5-shot settings, highlighting synthetic data as a scalable path for GFSS with reliable masks and precise boundaries.

Problem

• Generalized Few-shot Semantic Segmentation
• Novel-class Coverage
• Synthetic Images
• Pseudo-label Noise
• Mask Reliability

Innovation

• Generalized Few-shot Semantic Segmentation
• Diffusion-based Image Synthesis
• Prompt Bank Deduplication
• Support-guided Pseudo-labeling
• Boundary-aware SAM Refinement