Anomaly Detection On Mvtec Ad
Metrics
Detection AUROC
Results
Performance results of various models on this benchmark
Comparison Table
| Model Name | Detection AUROC |
|---|---|
| self-supervised-predictive-convolutional | 96.1 |
| efficientad-accurate-visual-anomaly-detection | 99.1 |
| towards-total-recall-in-industrial-anomaly | 99.6 |
| target-before-shooting-accurate-anomaly | 99.4 |
| fastflow-unsupervised-anomaly-detection-and | 99.4 |
| target-before-shooting-accurate-anomaly | - |
| winclip-zero-few-shot-anomaly-classification | 93.1 |
| dsr-a-dual-subspace-re-projection-network-for | 98.2 |
| attention-guided-anomaly-detection-and | - |
| deep-feature-selection-for-anomaly-detection | 96.6 |
| reconpatch-contrastive-patch-representation | - |
| efficientad-accurate-visual-anomaly-detection | 98.7 |
| focus-your-distribution-coarse-to-fine-non | 97.7 |
| uninformed-students-student-teacher-anomaly | - |
| probabilistic-distance-based-outlier | 98.2 |
| unsupervised-two-stage-anomaly-detection | 90 |
| a-unified-anomaly-synthesis-strategy-with | 99.9 |
| memseg-a-semi-supervised-method-for-image | 99.56 |
| cutpaste-self-supervised-learning-for-anomaly | - |
| template-guided-hierarchical-feature | 99.2 |
| attention-guided-anomaly-detection-and | - |
| iterative-energy-based-projection-on-a-normal-1 | - |
| self-supervised-context-learning-for-visual | 95.81 |
| sam-lad-segment-anything-model-meets-zero | 98.4 |
| target-before-shooting-accurate-anomaly | - |
| uninformed-students-student-teacher-anomaly | - |
| destseg-segmentation-guided-denoising-student | 98.6 |
| explainable-deep-one-class-classification | - |
| anomaly-detection-with-conditioned-denoising | 99.8 |
| padim-a-patch-distribution-modeling-framework | - |
| reconpatch-contrastive-patch-representation | 99.72 |
| unsupervised-anomaly-detection-and | 93.4 |
| registration-based-few-shot-anomaly-detection | 88.2 |
| student-teacher-feature-pyramid-matching-for | 95.5 |
| mocca-multi-layer-one-class-classification | 87.5 |
| winclip-zero-few-shot-anomaly-classification | 95.2 |
| diversity-measurable-anomaly-detection | 99.5 |
| winclip-zero-few-shot-anomaly-classification | 94.4 |
| padim-a-patch-distribution-modeling-framework | 97.9 |
| registration-based-few-shot-anomaly-detection | 92.7 |
| anomaly-localization-by-modeling-perceptual | - |
| collaborative-discrepancy-optimization-for-1 | - |
| anomaly-detection-via-reverse-distillation | 98.5 |
| image-anomaly-detection-and-localization-with | 99.56 |
| draem-a-discriminatively-trained | 98.0 |
| simplenet-a-simple-network-for-image-anomaly | 99.6 |
| reconstructed-student-teacher-and | 98.7 |
| unsupervised-anomaly-detection-and | 92.6 |
| modeling-the-distribution-of-normal-data-in | 95.8 |
| fair-frequency-aware-image-restoration-for | 98.6 |
| same-same-but-differnet-semi-supervised | 94.9 |
| towards-efficient-pixel-labeling-for | 99.7 |
| efficient-anomaly-detection-with-budget | 99.6 |
| adaclip-adapting-clip-with-hybrid-learnable | 89.2 |
| sub-image-anomaly-detection-with-deep-pyramid | 85.5 |
| unsupervised-image-anomaly-detection-and | 97.5 |
| target-before-shooting-accurate-anomaly | 99.7 |
| anomalydino-boosting-patch-based-few-shot | 96.6 |
| anomaly-detection-of-defect-using-energy-of | 95.1 |
| uninformed-students-student-teacher-anomaly | - |
| padim-a-patch-distribution-modeling-framework | 95.3 |
| transfusion-a-transparency-based-diffusion | 99.4 |
| reconpatch-contrastive-patch-representation | 99.62 |
| progressive-boundary-guided-anomaly-synthesis | 99.8 |
| reconstruction-from-edge-image-combined-with | 97.8 |
| informative-knowledge-distillation-for-image | - |
| anomalyclip-object-agnostic-prompt-learning | 91.5 |
| Model 68 | 99.5 |
| asymmetric-student-teacher-networks-for | 99.2 |
| target-before-shooting-accurate-anomaly | - |
| glad-towards-better-reconstruction-with | 99.3 |
| musc-zero-shot-industrial-anomaly | 97.8 |
| anomaly-detection-using-normalizing-flow | 98.85 |
| exploring-intrinsic-normal-prototypes-within | 99.8 |
| incremental-self-supervised-learning-based-on | 97.6 |
| multi-scale-patch-based-representation | 98.1 |
| unsupervised-anomaly-detection-and | 93.4 |
| semi-orthogonal-embedding-for-efficient | - |
| zero-shot-anomaly-detection-via-batch-1 | 85.8 |
| cflow-ad-real-time-unsupervised-anomaly | 98.26 |
| altub-alternating-training-method-to-update | - |
| reconstruction-by-inpainting-for-visual | 91.7 |
| industrial-anomaly-detection-with-domain | 98.4 |
| msflow-multi-scale-flow-based-framework-for | 99.7 |
| anoseg-anomaly-segmentation-network-using | 96 |
| position-encoding-enhanced-feature-mapping | - |
| vcp-clip-a-visual-context-prompting-model-for | - |
| inpainting-transformer-for-anomaly-detection | 95.0 |
| towards-total-recall-in-industrial-anomaly | 99.2 |
| exploring-dual-model-knowledge-distillation | 96.2 |
| grid-based-continuous-normal-representation | 99.4 |
| registration-based-few-shot-anomaly-detection | 91.2 |
| anomalydino-boosting-patch-based-few-shot | 97.7 |
| reconpatch-contrastive-patch-representation | 99.71 |
| anomaly-detection-via-self-organizing-map | 97.9 |
| anomaly-detection-of-defect-using-energy-of | 85.61 |
| anomalydino-boosting-patch-based-few-shot | 99.5 |
| learning-and-evaluating-representations-for-1 | 86.3 |
| two-stage-coarse-to-fine-image-anomaly | 98.0 |
| fully-convolutional-cross-scale-flows-for | 98.7 |
| attention-guided-anomaly-detection-and | - |
| cutpaste-self-supervised-learning-for-anomaly | 95.2 |
| altub-alternating-training-method-to-update | 99.4 |
| can-i-trust-my-anomaly-detection-system-a | 90 |
| generating-and-reweighting-dense-contrastive | 98.7 |
| self-supervised-predictive-convolutional | 98.9 |
| image-anomaly-detection-and-localization-with | 99.63 |
| excision-and-recovery-enhancing-surface | 94.2 |
| a-zero-few-shot-anomaly-classification-and | 86.1 |
| reconpatch-contrastive-patch-representation | 99.56 |
| self-supervised-out-of-distribution-detection-1 | 97.2 |
| cutpaste-self-supervised-learning-for-anomaly | 96.1 |
| omni-frequency-channel-selection | 98.3 |
| revisiting-reverse-distillation-for-anomaly | 99.44 |
| self-supervised-masked-convolutional | 98.7 |
| towards-total-recall-in-industrial-anomaly | 95.4 |
| n-pad-neighboring-pixel-based-industrial | 99.37 |
| explainable-deep-one-class-classification | - |
| unlocking-the-potential-of-reverse | 99.2 |
| 2408-03143 | 98.4 |
| self-supervised-masked-convolutional | 97.7 |
| winclip-zero-few-shot-anomaly-classification | 91.8 |
| multi-scale-feature-reconstruction-network | 98.4 |
| efficientad-accurate-visual-anomaly-detection | 99.8 |
| dfr-deep-feature-reconstruction-for | 93.8 |
| pyramidflow-high-resolution-defect | - |
| industrial-image-anomaly-localization-based | 93.1 |
| patch-svdd-patch-level-svdd-for-anomaly | 92.1 |
| realnet-a-feature-selection-network-with | 99.6 |
| target-before-shooting-accurate-anomaly | 99.7 |
| mean-shifted-contrastive-loss-for-anomaly | 87.2 |
| cfa-coupled-hypersphere-based-feature | 99.3 |
| learning-and-evaluating-representations-for-1 | 86.5 |
| dinomaly-the-less-is-more-philosophy-in-multi | 99.60 |
| anomalydino-boosting-patch-based-few-shot | 96.9 |
| fapm-fast-adaptive-patch-memory-for-real-time | 99 |
| Model 137 | 99.5 |
| produce-once-utilize-twice-for-anomaly | 99.5 |
| anomaly-detection-of-defect-using-energy-of | 89.02 |
| pedenet-image-anomaly-localization-via-patch | 92.8 |
| remembering-normality-memory-guided-knowledge | 99.6 |
| pyramidflow-high-resolution-defect | - |
| dinomaly-the-less-is-more-philosophy-in-multi | 99.77 |
| attention-guided-anomaly-detection-and | - |