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

2025-07-07

AI Summary

In industrial visual inspection, high-precision defect detection suffers from heavy reliance on large-scale annotated data and severe noise in pseudo-labels. To address these challenges, this paper proposes a semi-supervised framework integrating conditional diffusion models with CLIP-guided cross-modal alignment. The method employs a two-stage collaborative training and stepwise joint optimization strategy: first generating multi-scale pseudo-defect images from limited labeled samples, then leveraging CLIP’s semantic features for noise-aware pseudo-label filtering to significantly enhance pseudo-label quality. By innovatively combining generative modeling with cross-modal semantic alignment, the approach effectively mitigates label contamination. Evaluated on the NEU-DET dataset, our method achieves 75.1% mAP@0.5 using only 40% of the annotated data; under the same labeling budget, it further attains 78.4% mAP@0.5—demonstrating substantial improvements in both data efficiency and detection accuracy.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Semi-Supervised Learning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Semantics and Knowledge: Representation, semantic annotation, enhancement, enrichments, access and/or integration of a variety of data on the Web
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

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.

Problem

• Improves defect detection in industrial quality inspection
• Reduces reliance on labeled data via semi-supervised learning
• Enhances robustness with diffusion models and noise filtering

Innovation

• Semi-supervised framework with conditional diffusion
• CLIP-guided noise filtering for label contamination
• Two-stage training with pseudo-label generation