Anomaly Detection On Mvtec Loco Ad

Metrics

Avg. Detection AUROC

Detection AUROC (only logical)

Detection AUROC (only structural)

Segmentation AU-sPRO (until FPR 5%)

Results

Performance results of various models on this benchmark

Model Name	Avg. Detection AUROC	Detection AUROC (only logical)	Detection AUROC (only structural)	Segmentation AU-sPRO (until FPR 5%)	Paper Title	Repository
SPADE	68.9	70.9	66.8	45.1	Sub-Image Anomaly Detection with Deep Pyramid Correspondences
LADMIM	86.0	83.1	90.3	-	LADMIM: Logical Anomaly Detection with Masked Image Modeling in Discrete Latent Space	-
SLSG	90.3	89.6	91.4	67.3	SLSG: Industrial Image Anomaly Detection by Learning Better Feature Embeddings and One-Class Classification	-
CSAD	95.3	96.7	94.0	-	CSAD: Unsupervised Component Segmentation for Logical Anomaly Detection
DADF	83.7	79.2	88.2	67.4	Visual Anomaly Detection via Dual-Attention Transformer and Discriminative Flow	-
PatchCore	80.3	75.8	-	39.7	Towards Total Recall in Industrial Anomaly Detection
AST	-	79.7	87.1	42.7	Asymmetric Student-Teacher Networks for Industrial Anomaly Detection
SINBAD	86.8	88.9	84.7	-	Set Features for Fine-grained Anomaly Detection
PSAD	94.9	98.1	91.6	-	Few Shot Part Segmentation Reveals Compositional Logic for Industrial Anomaly Detection
ComAD+AST	89.8	90.1	89.4	-	Component-aware anomaly detection framework for adjustable and logical industrial visual inspection
Student-Teacher	77.3	66.4	88.3	-	Uninformed Students: Student-Teacher Anomaly Detection with Discriminative Latent Embeddings
PatchCore Ensemble	79.4	71.0	87.7	36.5	Towards Total Recall in Industrial Anomaly Detection
DSR	82.6	75.0	90.2	58.5	DSR -- A dual subspace re-projection network for surface anomaly detection
MNAD	65.1	60.0	70.2	33.9	Learning Memory-guided Normality for Anomaly Detection
HETMM	88.1	83.2	92.9	-	Hard-normal Example-aware Template Mutual Matching for Industrial Anomaly Detection
SINBAD Ens	88.3	91.2	85.5	-	Set Features for Anomaly Detection
DSKD	84.0	81.2	86.9	73.0	Contextual Affinity Distillation for Image Anomaly Detection	-
f-AnoGAN	64.2	65.8	62.7	33.4	f-AnoGAN: Fast Unsupervised Anomaly Detection with Generative Adversarial Networks
PUAD-S	93.1	92.0	94.1	-	PUAD: Frustratingly Simple Method for Robust Anomaly Detection
SINBAD+EfficientAD	94.2	95.8	94.2	-	Set Features for Anomaly Detection

0 of 40 row(s) selected.