Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
143 tokens/sec
GPT-4o
7 tokens/sec
Gemini 2.5 Pro Pro
46 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
38 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

REB: Reducing Biases in Representation for Industrial Anomaly Detection (2308.12577v2)

Published 24 Aug 2023 in cs.CV and cs.AI

Abstract: Existing representation-based methods usually conduct industrial anomaly detection in two stages: obtain feature representations with a pre-trained model and perform distance measures for anomaly detection. Among them, K-nearest neighbor (KNN) retrieval-based anomaly detection methods show promising results. However, the features are not fully exploited as these methods ignore domain bias of pre-trained models and the difference of local density in feature space, which limits the detection performance. In this paper, we propose Reducing Biases (REB) in representation by considering the domain bias and building a self-supervised learning task for better domain adaption with a defect generation strategy (DefectMaker) that ensures a strong diversity in the synthetic defects. Additionally, we propose a local-density KNN (LDKNN) to reduce the local density bias in the feature space and obtain effective anomaly detection. The proposed REB method achieves a promising result of 99.5\% Im.AUROC on the widely used MVTec AD, with smaller backbone networks such as Vgg11 and Resnet18. The method also achieves an impressive 88.8\% Im.AUROC on the MVTec LOCO AD dataset and a remarkable 96.0\% on the BTAD dataset, outperforming other representation-based approaches. These results indicate the effectiveness and efficiency of REB for practical industrial applications. Code:https://github.com/ShuaiLYU/REB.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (71)
  1. Trajectory-based surveillance analysis: A survey. IEEE Transactions on Circuits and Systems for Video Technology 29, 1985–1997. doi:https://doi.org/10.1109/TCSVT.2018.2857489.
  2. Variational autoencoder based anomaly detection using reconstruction probability URL: https://api.semanticscholar.org/CorpusID:36663713.
  3. Optics: Ordering points to identify the clustering structure. ACM Sigmod record 28, 49–60. doi:https://doi.org/10.1145/304181.304187.
  4. Deep nearest neighbor anomaly detection. arXiv preprint arXiv:2002.10445 doi:https://doi.org/10.48550/arXiv.2002.10445.
  5. Beyond dents and scratches: Logical constraints in unsupervised anomaly detection and localization. International Journal of Computer Vision 130, 947–969. doi:https://doi.org/10.1007/s11263-022-01578-9.
  6. Mvtec ad — a comprehensive real-world dataset for unsupervised anomaly detection , 9584–9592doi:https://doi.org/10.1109/cvpr.2019.00982.
  7. Uninformed students: Student-teacher anomaly detection with discriminative latent embeddings , 4183–4192doi:https://doi.org/10.1109/cvpr42600.2020.00424.
  8. Lof: identifying density-based local outliers , 93–104doi:https://doi.org/10.1145/342009.335388.
  9. Nearest-neighbor clutter removal for estimating features in spatial point processes. Journal of the American Statistical Association 93, 577–584. doi:https://doi.org/10.1080/01621459.1998.10473711.
  10. Informative knowledge distillation for image anomaly segmentation. Knowledge-Based Systems 248, 108846. doi:https://doi.org/10.1016/j.knosys.2022.108846.
  11. Anomaly detection of defects on concrete structures with the convolutional autoencoder. Advanced Engineering Informatics 45, 101105. doi:https://doi.org/10.1016/j.aei.2020.101105.
  12. Coresets, sparse greedy approximation, and the frank-wolfe algorithm. ACM Transactions on Algorithms (TALG) 6, 1–30. doi:https://doi.org/10.1145/1824777.1824783.
  13. Sub-image anomaly detection with deep pyramid correspondences. arXiv preprint arXiv:2005.02357 doi:https://doi.org/10.48550/arXiv.2005.02357.
  14. Padim: a patch distribution modeling framework for anomaly detection and localization , 475–489doi:https://doi.org/10.1007/978-3-030-68799-1_35.
  15. Imagenet: A large-scale hierarchical image database , 248–255doi:10.1109/CVPR.2009.5206848.
  16. Improved regularization of convolutional neural networks with cutout. arXiv preprint arXiv:1708.04552 doi:https://doi.org/10.48550/arXiv.1708.04552.
  17. Understanding receiver operating characteristic (roc) curves. Canadian Journal of Emergency Medicine 8, 19–20.
  18. Algorithms for rational bézier curves. Computer-aided design 15, 73–77. doi:https://doi.org/10.1016/0010-4485(83)90171-9.
  19. Memorizing normality to detect anomaly: Memory-augmented deep autoencoder for unsupervised anomaly detection , 1705–1714doi:https://doi.org/10.1109/iccv.2019.00179.
  20. Cflow-ad: Real-time unsupervised anomaly detection with localization via conditional normalizing flows , 98–107doi:https://doi.org/10.1109/wacv51458.2022.00188.
  21. Template-guided hierarchical feature restoration for anomaly detection , 6447–6458doi:https://doi.org/10.1109/iccv51070.2023.00593.
  22. Deep residual learning for image recognition , 770–778doi:https://doi.org/10.1109/cvpr.2016.90.
  23. Reconpatch: Contrastive patch representation learning for industrial anomaly detection , 2052–2061.
  24. A survey on contrastive self-supervised learning. Technologies 9, 2. doi:https://doi.org/10.3390/technologies9010002.
  25. A masked reverse knowledge distillation method incorporating global and local information for image anomaly detection. Knowledge-Based Systems 280, 110982. doi:https://doi.org/10.2139/ssrn.4354294.
  26. Loop: local outlier probabilities , 1649–1652.
  27. Outlier detection with kernel density functions 7, 61–75. doi:https://doi.org/10.1007/978-3-540-73499-4_6.
  28. Pyramidflow: High-resolution defect contrastive localization using pyramid normalizing flow , 14143–14152.
  29. Cutpaste: Self-supervised learning for anomaly detection and localization , 9664–9674doi:https://doi.org/10.1109/cvpr46437.2021.00954.
  30. Visual saliency detection based on multiscale deep cnn features. IEEE transactions on image processing 25, 5012–5024. doi:10.1109/TIP.2016.2602079.
  31. Superpixel masking and inpainting for self-supervised anomaly detection URL: https://api.semanticscholar.org/CorpusID:221669098.
  32. Unsupervised image anomaly detection and localization in industry based on self-updated memory and center clustering. IEEE Transactions on Instrumentation and Measurement doi:10.1109/TIM.2023.3271754.
  33. Simplenet: A simple network for image anomaly detection and localization , 20402–20411doi:https://doi.org/10.1109/cvpr52729.2023.01954.
  34. Automatic fabric defect detection with a multi-scale convolutional denoising autoencoder network model. Sensors 18, 1064. doi:https://doi.org/10.3390/s18041064.
  35. Vt-adl: A vision transformer network for image anomaly detection and localization , 01–06doi:https://doi.org/10.1109/isie45552.2021.9576231.
  36. Understanding auc-roc curve. Towards data science 26, 220–227.
  37. Deep learning for anomaly detection: A review. ACM computing surveys (CSUR) 54, 1–38. doi:https://doi.org/10.1145/3439950.
  38. Learning memory-guided normality for anomaly detection , 14372–14381doi:https://doi.org/10.1109/cvpr42600.2020.01438.
  39. Modeling the distribution of normal data in pre-trained deep features for anomaly detection , 6726–6733doi:https://doi.org/10.1109/icpr48806.2021.9412109.
  40. Towards total recall in industrial anomaly detection , 14298–14308doi:https://doi.org/10.1109/cvpr52688.2022.01392.
  41. Same same but different: Semi-supervised defect detection with normalizing flows , 1907–1916doi:https://doi.org/10.1109/wacv48630.2021.00195.
  42. Fully convolutional cross-scale-flows for image-based defect detection , 1088–1097doi:https://doi.org/10.1109/wacv51458.2022.00189.
  43. Asymmetric student-teacher networks for industrial anomaly detection , 2592–2602doi:https://doi.org/10.48550/arXiv.2210.07829.
  44. Unsupervised anomaly detection with generative adversarial networks to guide marker discovery , 146–157doi:https://doi.org/10.1007/978-3-319-59050-9_12.
  45. Natural synthetic anomalies for self-supervised anomaly detection and localization , 474–489doi:https://doi.org/10.1007/978-3-031-19821-2_27.
  46. Dbscan revisited, revisited: why and how you should (still) use dbscan. ACM Transactions on Database Systems (TODS) 42, 1–21. doi:https://doi.org/10.1145/3068335.
  47. Exploiting epistemic uncertainty of anatomy segmentation for anomaly detection in retinal oct. IEEE Transactions on Medical Imaging 39, 87–98. doi:https://doi.org/10.1109/TMI.2019.2919951.
  48. Unsupervised identification of disease marker candidates in retinal oct imaging data. IEEE Transactions on Medical Imaging 38, 1037–1047. doi:https://doi.org/10.1109/TMI.2018.2877080.
  49. Active learning for convolutional neural networks: A core-set approach. arXiv preprint arXiv:1708.00489 doi:https://doi.org/10.48550/arXiv.1708.00489.
  50. Learning and evaluating representations for deep one-class classification. arXiv preprint arXiv:2011.02578 doi:https://doi.org/10.48550/arXiv.2011.02578.
  51. Similarity measures for occlusion, clutter, and illumination invariant object recognition , 148–154doi:https://doi.org/10.1007/3-540-45404-7_20.
  52. Detecting outliers with foreign patch interpolation. arXiv preprint arXiv:2011.04197 doi:https://doi.org/10.48550/arXiv.2011.04197.
  53. Deep learning for unsupervised anomaly localization in industrial images: A survey. IEEE Transactions on Instrumentation and Measurement 71, 1–21. doi:10.1109/TIM.2022.3196436.
  54. Distance-based anomaly detection for industrial surfaces using triplet networks , 0372–0377doi:https://doi.org/10.1109/iemcon51383.2020.9284921.
  55. Revisiting reverse distillation for anomaly detection , 24511–24520doi:https://doi.org/10.1109/cvpr52729.2023.02348.
  56. Autoencoder-based anomaly detection for surface defect inspection. Advanced Engineering Informatics 48, 101272. doi:https://doi.org/10.1016/j.aei.2021.101272.
  57. Set features for fine-grained anomaly detection. arXiv preprint arXiv:2302.12245 doi:https://doi.org/10.48550/arXiv.2302.12245.
  58. Industrial image anomaly localization based on gaussian clustering of pretrained feature. IEEE Transactions on Industrial Electronics 69, 6182–6192. doi:https://doi.org/10.1109/TIE.2021.3094452.
  59. Edn: Salient object detection via extremely-downsampled network. IEEE Transactions on Image Processing 31, 3125–3136. doi:10.1109/TIP.2022.3164550.
  60. Learning semantic context from normal samples for unsupervised anomaly detection 35, 3110–3118. doi:https://doi.org/10.1609/aaai.v35i4.16420.
  61. Fastflow: Unsupervised anomaly detection and localization via 2d normalizing flows. arXiv preprint arXiv:2111.07677 doi:https://doi.org/10.48550/arXiv.2111.07677.
  62. Wide residual networks. arXiv preprint arXiv:1605.07146 doi:https://doi.org/10.48550/arXiv.1605.07146.
  63. Draem-a discriminatively trained reconstruction embedding for surface anomaly detection , 8330–8339doi:https://doi.org/10.1109/iccv48922.2021.00822.
  64. Reconstruction by inpainting for visual anomaly detection. Pattern Recognition 112, 107706. doi:https://doi.org/10.1016/j.patcog.2020.107706.
  65. Defgan: Defect detection gans with latent space pitting for high-speed railway insulator. IEEE Transactions on Instrumentation and Measurement 70, 1–10. doi:https://doi.org/10.1109/tim.2020.3038008.
  66. mixup: Beyond empirical risk minimization. arXiv preprint arXiv:1710.09412 doi:https://doi.org/10.48550/arXiv.1710.09412.
  67. A new local distance-based outlier detection approach for scattered real-world data , 813–822doi:https://doi.org/10.1007/978-3-642-01307-2_84.
  68. Destseg: Segmentation guided denoising student-teacher for anomaly detection , 3914–3923doi:https://doi.org/10.1109/cvpr52729.2023.00381.
  69. A surface defect detection method based on positive samples , 473–481doi:https://doi.org/10.1007/978-3-319-97310-4_54.
  70. Focus your distribution: Coarse-to-fine non-contrastive learning for anomaly detection and localization , 1–6doi:https://doi.org/10.1109/icme52920.2022.9859925.
  71. Unsupervised anomaly localization using variational auto-encoders , 289–297doi:https://doi.org/10.1007/978-3-658-29267-6_43.
Citations (6)
View on Semantic Scholar

Summary

We haven't generated a summary for this paper yet.

Ai Sparkles Streamline Icon: https://streamlinehq.com Summarize Now
X Twitter Logo Streamline Icon: https://streamlinehq.com

Tweets