cs.CVAug 18, 2026

Continuity-Driven Representation Learning for Industrial Defect Detection

Authors: Minjong Kim, Hyun Jun Kim, Jeongrae Kim, Heeseung Shin, Changwon Lim

Organizations: Chung-Ang University Seoul, Republic of Korea

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Industrial defect detection differs from natural-image object detection because inspection images are captured under controlled conditions and contain large normal-dominant regions with repetitive structures. Defects therefore appear as localized disruptions of otherwise predictable patterns, while conventional detectors rely mainly on sparse bounding-box supervision, resulting in weakly constrained normal-region representations. We propose a continuity-driven representation regularization framework that exploits normal-dominant regions as dense auxiliary supervision. The framework introduces two detector-agnostic objectives: Multi-Continuity Loss, which combines 1D patch-sequence prediction and 2D masked spatial prediction, and Differencing Loss, which regularizes first-order feature variation and second-order curvature between neighboring patch embeddings. Both objectives are applied with box-derived region weighting to stabilize normal-region representations while preserving defect-related discontinuities. Experiments on two real-world industrial datasets and the public NEU-DET benchmark, using six detector architectures including YOLO-family models, MambaYOLO, and DETR, demonstrate consistent improvements over native detector baselines. In the full-data setting, the proposed regularizers improve average [email protected]:0.95 by up to 3.49 percentage points on Industrial Metal, 5.38 percentage points on MEA, and 5.03 percentage points on NEU-DET. Under limited-data conditions, the gains become more pronounced, with Differencing Loss achieving improvements of up to 21.07 percentage points in [email protected] and 8.23 percentage points in [email protected]:0.95 on NEU-DET using only 25% of the training data. These results suggest that continuity-driven regularization provides an effective prior for improving industrial defect detection, particularly when annotated data are scarce.

Figures & tables

Figure 1: Comparison between natural-image object detection and industrial defect detection. Industrial inspection images contain repetitive normal structures, where defects appear as localized disruptions of structural continuity.
Figure 2: Overall framework of the proposed continuity-driven representation regularization. Intermediate detector features are regularized using Multi-Continuity Loss or Differencing Loss with box-derived region weighting during training.
Figure 3: Industrial inspection examples used for qualitative illustration. Left: an Industrial Metal inspection image. Right: schematic illustration of the MEA inspection region and localized defect examples. The schematic is used only for visualization and is not used for training or evaluation.
Figure 4: Qualitative comparison on NEU-DET under 100% and 25% training-data settings. Differencing recovers scratch defects missed by the baseline, while the 25% setting also shows a detection on an ambiguous defect-like structure.
Appendix figures & tables3 assets

Supplementary material from the paper’s appendix.

Appendix

ScalePrecisionRecall[email protected][email protected]:0.95
0.5×0.5\times0.9160.9200.9750.707
0.7×0.7\times0.7910.8940.9330.637
0.9×0.9\times0.9050.8250.9360.618
1.0×1.0\times0.8930.9460.9650.701
1.1×1.1\times0.9250.8860.9450.651
1.3×1.3\times0.9340.8750.9710.723
Appendix
Table A1: Robustness to different bounding-box scales on Industrial Metal using YOLOv12 with Differencing Loss.
Figure B1: Qualitative comparison of feature representations from YOLOv12. The baseline model produces more dispersed feature responses, whereas the proposed continuity-driven regularization yields more localized activations around defect regions while preserving the underlying normal structures.
Figure B2: Continuity discrepancy visualization from intermediate YOLOv12 feature maps. The left column shows the baseline model trained with the standard detection loss, and the right column shows the continuity-regularized model trained with the proposed auxiliary objective. Continuity discrepancy maps visualize local violations of feature continuity and are used only as diagnostic visualizations, not as detector outputs. Strong responses may occur around defect-related disruptions as well as normal object boundaries or abrupt structural transitions.

Explore similar work

CardsList
Aug 24, 2026

What Remains Normal? Clean Images Miss Useful Near-Defect Normal Patches for Anomaly Detection

Normal-only industrial anomaly detectors use patches from clean training images as normal references or reconstruction targets. This assumes that clean patches are sufficient for the normal regions encountered at test time. We test that assumption directly. On MVTec AD, admitting ground-truth-normal patches from real defect images to a DINOv2 memory candidate pool raises pixel average precision (P-AP) from 73.34 to 76.95 while keeping the encoder, test-time score, and number of stored references fixed. Patches within two patch cells of the annotated defect recover 94.70% of this gain. We then ask whether useful patches of this kind can be exposed using clean training images alone. BoundarySupport inserts a procedural synthetic defect to alter surrounding context, excludes every token intersecting the nominal insertion or a detected RGB change, and learns only from pixel-preserved neighboring patches. Across three paired seeds, the same principle improves P-AP in all six memory and reconstruction settings across MVTec, VisA, and Real-IAD. Matched controls identify the altered-context feature itself as the useful normal evidence: with synthetic input or selected positions fixed, altered-context features outperform their clean-view counterparts as both reconstruction targets and memory references. On MVTec memory, the final score change is also spatially selective, with larger reductions on normal patches next to defects than on mid-distance or far-normal patches in all 15 categories. Code is publicly available at https://github.com/jw-chae/boundary_support.

Joongwon Chae, Runming Wang, Peiwu Qin

Institute of Biopharmaceutical and Health Engineering, Tsinghua Shenzhen International Graduate School, Tsinghua University, Shenzhen, China · RatelSoft · Guangdong Provincial Laboratory of Traditional Chinese Medicine, Hengqin, Guangdong, China

Industrial Anomaly DetectionReconstruction-Based Anomaly Detection
AskPDF
Sep 3, 2026

Neural-Collapse-guided Task-Free Continual Anomaly Detection

Recent years have witnessed growing interest in continual anomaly detection for industrial visual inspection. However, real-world manufacturing environments exhibit unpredictable shifts in data distributions, rendering task-dependent continual learning assumptions impractical. To address this limitation, we formulate industrial anomaly detection as a task-free continual learning problem and propose NC-TFAD, a neural-collapse-inspired, geometry-driven framework for learning from non-stationary data streams without task boundaries. NC-TFAD freezes a pretrained backbone and aligns streaming features to a simplex Equiangular Tight Frame (ETF) prototype space to stabilize representation geometry under non-stationary streams. To satisfy the NC-inspired geometric construction in the absence of real anomalies, we generate synthetic anomaly samples as auxiliary anchors during training. Building on this geometry, we further introduce inter- and intra-class regularization together with a Focal Neural Collapse Contrastive (FNCC) loss to suppress representation drift and improve normal-anomaly separability. Finally, a normal-patch-prototype-guided localization branch constructs calibrated patch-wise deviation maps from normal training samples and fuses them with a weak self-attention prior, producing anomaly heatmaps without pixel-level annotations. Extensive experiments on MVTec AD and VisA show that NC-TFAD consistently outperforms representative task-free continual learning methods adapted from general vision, as well as unified anomaly detection baselines, in both image-level detection and pixel-level localization under the task-free continual learning protocol. These results highlight that geometry-driven modeling offers an effective and robust solution for task-free continual anomaly detection in real-world industrial applications.

Xiaotong Kong, Chaoyang Song, Ziai Zhou +3

Key Laboratory of Measurement and Control of CSE, Ministry of Education, School of Automation, Southeast University, Nanjing 210096, China

Continual LearningIndustrial Anomaly DetectionAnomaly LocalizationAnomaly Detection
AskPDF
Oct 10, 2026

InCLAD: A Continual Learning Benchmark for Industrial Visual Anomaly Detection

Industrial anomaly detection benchmarks have advanced rapidly, yet they remain largely static and therefore underrepresent deployment settings in which inspection systems must be extended from one industrial component to another over time. In our setting, continual industrial anomaly detection is framed as a task-incremental problem: each component defines a task, while the learning objective remains semi-supervised. The goal extends beyond conventional anomaly detection, since models should preserve their performance on all components as they are sequentially exposed to new components. This framing is especially relevant because many existing industrial pipelines either assume fully supervised defect labels or study only isolated datasets, leaving open the question of how semi-supervised anomaly detectors behave across a sequence of heterogeneous components. We address this gap by organizing the benchmark as a sequence of related component-level tasks and by defining multiple scenarios, including random, easy-to-hard, and hard-to-easy. In addition to single-dataset continual scenarios, we propose challenging scenarios spanning across multiple datasets. We further evaluate continual performance from two complementary perspectives: image-level anomaly detection and pixel-level anomaly localization. By recasting industrial anomaly detection as a continual benchmark rather than a static train/test problem, our work establishes a concrete path toward studying knowledge retention, adaptation, and generalization in evolving industrial environments.

Łukasz Marcjan, Dorota Wilk-Kołodziejczyk, Roberto Corizzo

AGH University of Krakow, Krakow, Poland · American University, Washington, DC, USA

Industrial Anomaly DetectionAnomaly DetectionAnomaly Localization
AskPDF