P-SAM: SAM with Perceptual Parallel Prompt for Few-Shot Strip Steel Surface Defect Segmentation
En palabras de los autores
Few-shot semantic segmentation (FSS) of strip steel surface defects (SD) has posed significant challenges distinct from natural scenes. Unlike natural images, SD task exhibits unique characteristics including low local contrast, uneven illumination, and complex fine-grained texture patterns. Although recent methods based on Segment Anything Model (SAM) have shown promise in FSS on natural images by leveraging SAM's powerful pre-trained representations, these unique industrial characteristics of SD images lead to performance drop when directly applying SAM to industrial defect scenarios. In this paper, we propose a novel Perceptual Parallel Prompt (P) framework that empowers SAM, creating the P-SAM model to address these challenges through two core strategies. First, we develop a Perceptual-Optimized Encoding (POE) strategy that enhances local contrast and preserves critical texture details for SD segmentation. Second, we introduce the Parallel Prompt Generator (PPG) strategy that simultaneously generates both semantic and spatial prompts, enabling comprehensive guidance for SAM's decoder across varying images. Extensive experiments on three few-shot SD benchmarks demonstrate that P-SAM achieves state-of-the-art performance, with particularly notable improvements of 12.00% in mIoU on Surface Defects-4i dataset.
Apareció: lunes, 21 de septiembre. arXiv. Preprint, todavía sin revisión por pares.
Comentario de los autores: Accepted by ICME 2026, 6 pages, 3 figures. Corresponding authors: Anpeng Wang and Runmin Cong