# Localization and Pixel-Confidence Network for Surface Defect Segmentation

## TL;DR

This paper introduces a two-stage deep learning network to improve surface defect segmentation in industrial inspection by addressing imbalanced data and over-segmentation issues.

## Contribution

A novel two-stage network combining a Defect Localization Module and a Pixel Confidence Module for better segmentation accuracy.

## Key findings

- The proposed network achieved 1.58%±0.80% improvement in mPA on the Carbon Fabric Defect Dataset.
- It showed 2.66%±1.12% increase in mPA on the Magnetic Tile Defect Dataset.
- The method reduces over-segmentation and improves automated quality assurance in industrial settings.

## Abstract

Surface defect segmentation based on deep learning has been widely applied in industrial inspection. However, two major challenges persist in specific application scenarios: first, the imbalanced area distribution between defects and the background leads to degraded segmentation performance; second, fine gaps within defects are prone to over-segmentation. To address these issues, this study proposes a two-stage image segmentation network that integrates a Defect Localization Module and a Pixel Confidence Module. In the first stage, the Defect Localization Module performs a coarse localization of defect regions and embeds the resulting feature vectors into the backbone of the second stage. In the second stage, the Pixel Confidence Module captures the probabilistic distribution of neighboring pixels, thereby refining the initial predictions. Experimental results demonstrate that the improved network achieves gains of 1.58%±0.80% in mPA, 1.35%±0.77% in mIoU on the self-built Carbon Fabric Defect Dataset and 2.66%±1.12% in mPA, 1.44%±0.79% in mIoU on the public Magnetic Tile Defect Dataset compared to the other network. These enhancements translate to more reliable automated quality assurance in industrial production environments.