Industrial anomaly detection improved by defect grounded reasoning in latent space
Industrial Anomaly Detection via Defect-Grounded Reasoning in Visual Latent Space
Computer Vision and Pattern Recognition
Summary
Finding tiny flaws in industrial images can be tricky. The authors created a new method that first looks at the whole image to understand it, then zooms in on the important parts in a smarter way inside the computer’s image understanding system. They also built a big set of example images and questions to help train and test this method. This approach works better than similar ones without needing extra tools or references.
What this means in practice
- •For industrial quality control teams: Improve automation of defect detection by using a method that better understands and refines subtle anomalies directly from images without external tools.
- •For manufacturing ai developers: Develop smarter inspection systems incorporating global and local defect reasoning to provide more detailed explanations of anomalies in industrial parts.
Authors
Jaron Yeh, Yen-Wei Chang, Jiang Liu, Shao-Yuan Lo
Abstract
Industrial anomaly detection (IAD) is evolving beyond conventional detection and localization toward multimodal inspection systems that can describe, explain, and reason about fine-grained defects. Although recent multimodal large language model (MLLM)-based methods improve anomaly understanding through textual reasoning and visual guidance, they face two limitations in fine-grained inspection. First, their visual refinement often requires iteratively revisiting local image regions or augmenting with additional tools. Second, the resulting local defect evidence may not be reliably preserved throughout subsequent reasoning. To address these, we propose Anomaly-LR, a defect-grounded latent reasoning framework that first forms a global understanding of the input and then progressively refines anomaly-relevant representations directly in the visual latent space. We further construct IAD-LR-22K, the first IAD instruction dataset designed for latent reasoning, containing 22,228 image-question instances from 4,523 industrial images, with global textual reasoning traces and region-level visual annotations. Extensive experiments show that Anomaly-LR achieves state-of-the-art performance among comparable-scale methods across multiple IAD benchmarks, without requiring external references or tools. The code and data will be released at https://github.com/Yen666/Anomaly-LR.