Semi-Supervised Defect Detection via Conditional Diffusion and CLIP-Guided Noise Filtering

TL;DR

This paper presents a semi-supervised defect detection framework using conditional diffusion and CLIP-guided noise filtering, achieving high accuracy with less labeled data in industrial inspection tasks.

Contribution

It introduces a novel semi-supervised approach combining conditional diffusion and CLIP-based noise filtering for defect detection, reducing labeling requirements.

Findings

Achieves 78.4% mAP@0.5 with full labeled data

Reaches 75.1% mAP@0.5 with only 40% labeled data

Demonstrates superior data efficiency over traditional supervised methods

Abstract

In the realm of industrial quality inspection, defect detection stands as a critical component, particularly in high-precision, safety-critical sectors such as automotive components aerospace, and medical devices. Traditional methods, reliant on manual inspection or early image processing algorithms, suffer from inefficiencies, high costs, and limited robustness. This paper introduces a semi-supervised defect detection framework based on conditional diffusion (DSYM), leveraging a two-stage collaborative training mechanism and a staged joint optimization strategy. The framework utilizes labeled data for initial training and subsequently incorporates unlabeled data through the generation of pseudo-labels. A conditional diffusion model synthesizes multi-scale pseudo-defect samples, while a CLIP cross-modal feature-based noise filtering mechanism mitigates label contamination. Experimental results on the NEU-DET dataset demonstrate a 78.4% mAP@0.5 with the same amount of labeled data as traditional supervised methods, and 75.1% mAP@0.5 with only 40% of the labeled data required by the original supervised model, showcasing significant advantages in data efficiency. This research provides a high-precision, low-labeling-dependent solution for defect detection in industrial quality inspection scenarios. The work of this article has been open-sourced at https://github.com/cLin-c/Semisupervised-DSYM.