A unified human-in-the-loop framework for manufactured-part inspection that combines image annotation, AI-assisted defect detection, and an integrated validation engine is developed, replacing a prior manual visual inspection and documentation workflow.
Abstract
Automated anomaly detection methods often report strong performance on curated academic benchmarks, but their behavior under real-world industrial conditions is less clear. In this work, we evaluate 19 unsupervised anomaly detection models on the BowTie dataset, a challenging manufacturing dataset with reflective surfaces, subtle defects, and profile-specific variation. In contrast to benchmark results, we observe that model performance is less stable than typically reported on standard benchmarks such as MVTec AD, highly sensitive to preprocessing, and inconsistent across conditions, with no single approach emerging as uniformly robust; a consensus audit further indicates that nominal-data quality affects deployment. Motivated by these findings, we developed and initially deployed a unified human-in-the-loop framework for manufactured-part inspection that combines image annotation, AI-assisted defect detection, and an integrated validation engine, replacing a prior manual visual inspection and documentation workflow. The system supports heatmap-guided defect review, SAM-refined candidate regions for inspector acceptance, rejection, or boundary adjustment, mask evaluation where annotations exist, and review history for inspector consistency and onboarding. Together, the results highlight the gap between benchmark performance and deployment reality, and provide a practical framework for addressing it.
This work benchmarks four state-of-the-art unsupervised anomaly detection methods CFA, DFM, FRE, and PatchCore for automated PCB inspection under a strict one-class learning protocol. All models were trained exclusively on nominal samples and evaluated using default configurations to reflect realistic industrial deployment. Experiments on the MVTec AD Transistor dataset and a high-density Raspberry Pi PCB dataset assessed imageand pixel-level performance using AUROC, F1-Score, and AUPRO. PatchCore achieved the strongest image-level discrimination, CFA provided the most balanced detection–localization trade-off, FRE showed sensitivity to localized structural defects, and DFM performed well globally but with weaker boundary precision. Results demonstrate the practical viability of unsupervised deep anomaly detection for modern AOI systems in dynamic manufacturing environments.
Dário Alef Barros Lima, Karen Letícia Santana da Silva, Willian Do Nascimento Severiano et al.· Anais do LIII Seminário Inte...· 0 citations
LLM-Detector is proposed, a framework that utilizes the in-context learning capacity of LLMs for structured, prompt-conditioned scoring synthesis, enabling LLMs to derive anomaly detection logic from structured normal-state knowledge.
Tu Nguyen, Dang Nguyen, Thuc Duy Le et al.· 0 citations
Industrial anomaly detection is a critical component of modern manufacturing. Most traditional unsupervised methods rely on modelling normal feature distributions, inherently limiting generalization to unknown categories. To improve generalizability, some recent methods incorporate vision-language models (VLMs) for zero-shot detection via text prompts. However, we observe that reasoning-oriented post-training can cause anomaly discrimination to collapse, with some fine-tuned models performing worse than their base VLMs. Existing methods also provide only textual decisions or coarse boxes, without pixel-level segmentation. A more explicit detection principle comes from human inspection: anomalies are identified by comparing a query image with a defect-free reference. Inspired by this, we propose InspectorGPT, a VLM framework centered on comparative reasoning. Given a normal reference and a query image, InspectorGPT compares them to identify discrepancies and perform multiple inspection tasks with detailed reasoning. We internalize this capability through Chain-of-Thought (CoT) fine-tuning and Group Relative Policy Optimization (GRPO) with tailored, verifiable rewards. We further introduce InspectorGPT-Seg for pixel-level anomaly masks. Segmentation supervision improves anomaly discrimination but weakens semantic reasoning, while joint training fails to balance them. We therefore train the two branches separately and combine them through task-vector fusion. Extensive experiments demonstrate superior multi-dimensional performance and generalization to unseen benchmarks, validating comparative reasoning for comprehensive industrial inspection.
Weifeng Chen, Hong-Hao Zhang, Zhiyuan You et al.· 0 citations
Recent advances in anomaly detection (AD) for industrial inspection have pushed performance on standard benchmarks toward saturation. However, strong benchmark performance does not necessarily translate to real-world deployment, as these benchmarks are primarily collected under controlled acquisition conditions. Changes in illumination, background, viewpoint, and other environmental factors can shift normal samples away from the learned normal distribution and cause false anomaly responses. We address AD under such distribution shifts by explicitly modeling nuisance variation from changing imaging conditions in feature space. Without anomaly labels or target-domain data, our Nuisance-Filtered Anomaly Detection (NFAD) framework estimates a nuisance subspace from matched feature displacements induced by content-preserving perturbations and suppresses its contribution to anomaly residuals at inference. The same subspace supports two complementary branches: full projection for image-level detection and selective suppression for pixel-level localization, preserving evidence of localized defects. On AeBAD-S, a benchmark specifically designed for AD under acquisition shifts, NFAD achieves 91.0\% image-level AUROC, establishing a new state of the art. Notably, this robustness does not come at the expense of conventional AD performance: NFAD remains competitive on standard benchmarks that do not explicitly evaluate distribution shift, including VisA, Real-IAD, and MVTec AD. These results show that explicitly suppressing such nuisance variation improves AD under distribution shift while preserving strong performance in standard settings.
Dat Cao, Son T. Nghiem, Phan Nguyen et al.· 0 citations
Automation, AI, and modern monitoring equipment has made the dream of surveillance staff having a near-omniscient view of oilfield issues closer to reality. This paper tackles a segment of this vision and presents a low-code well surveillance workflow developed to reduce the manual monitoring by well performance specialists and production engineers while extending surveillance coverage across gas lift wells. The proof-of-concept system was built by Occidental using Microsoft Power Automate, Power Apps, SharePoint, OneDrive, Azure OCR, and multimodal large language model (LLM) calls.
The workflow captures well images, extracts structured information, prompts an LLM to identify abnormal operating conditions and propose follow-up actions, and then presents those outputs in a user-facing application alongside the original surveillance view. The prototype established a complete end-to-end path from data capture to anomaly review and user feedback. In testing, the system demonstrated that a low-code architecture can support daily surveillance screening, highlight wells that merit attention, and place draft recommendations directly in front of users.
This work is not presented as a final enterprise architecture. Instead, it documents the decisions, tradeoffs, performance metrics comparing different LLMs, and lessons from an accelerated proof of concept effort. The results show that low-code tools can be used to create a practical, AI-assisted surveillance workflow, create a testing ground for multimodal AI in production operations, and provide a bridge toward future model-driven surveillance systems that rely on structured data and purpose-built machine learning models.
Ivan Tanakov, Emmanuel Zoubovsky, Ivan Berry et al.· Americas· 0 citations
Anomaly detection methods often have uncertain behavior with respect to samples near the distribution boundary, limiting their ability to anticipate future anomalies. This work introduces the concept of near-anomalies that, while not yet anomalous, lie close to the boundary and are likely to transition into anomalies in the near future. To address this, we propose an unsupervised method, named Christoffel-based ANomaly Anticipation for eaRly dIscovery (CANARI), which leverages the strong theoretical foundations of the Christoffel function to detect near-anomalies. The method is validated on industrial in-circuit testing data from printed circuit boards, with synthetically generated near-anomaly samples due to the lack of real-world data labeling. Experimental results show that CANARI outperforms the compared baselines that generally use a dual-threshold mechanism (one for anomalies and one for near-anomalies). It therefore provides a proactive solution for anticipating anomalies before they occur, offering a promising approach for resilience, predictive maintenance, and quality control.
Léa Billet, Louise Travé-Massuyès, E. Chanthery et al.· arXiv.org· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
