List your company

Anomaly Detection on MVTec AD (Per-Class AUROC Breakdown)

99.8Overall AUROCCPR*
73.38480.24287.193.958Nov 15, 2021Apr 6, 2022Aug 27, 2022Jan 17, 2023Jun 8, 2023Oct 29, 2023Mar 20, 2024

Evaluation results

MethodLinks
CPR*variant=*2023.0899.8—————————————————
CPR2023.0899.7—————————————————
SimpleNet2023.0899.6—————————————————
SimpleNetType=Ours2023.0399.599.799.710099.810010099.997.71001009998.299.710099.999.699.8
FastFlow2021.1199.410099.710010010010010010010010099.497.894.499.899.5——
FastFlow2022.1099.410099.710010010010010010010010099.497.894.499.899.5——
N-padCategory=Pre-trained Feature Based2022.1099.3799.2798.6710010099.5610099.5499.41001009897.410099.5899.34——
PatchCoreType=Embedding-based2023.0399.298.798.210098.799.210099.598.110010096.698.110010099.499.199
PatchCore2021.1199.198.798.210098.799.210099.598.110010096.698.110010098.8——
PatchCore2022.1099.198.798.610099.499.210099.598.11001009798.110010099.5——
PatchCoreCategory=Pre-trained Feature Based2022.109998.797.910098.99910099.497.8100100969799.710099.5——
SSPCABBase model=DRAEM2022.1098.998.210010010099.598.496.999.310010099.897.910092.9100——
SSPCAB2023.0898.9—————————————————
CS-Flow2022.1098.79910010010010099.899.197.199.699.198.697.691.999.399.7——
DeSTSeg2023.0898.6—————————————————
RLR2024.0398.699.799.810010098.910099.39710099.998.994.893.199.798.5——
PatchCore2023.0898.5—————————————————
RD2023.0898.5—————————————————
AltUBBase model=CFLOW-AD2022.1098.499.210010099.99910097.898.110099.59791.799.495.298.5——
HGADEvaluation Protocol=Unified case2024.0398.410099.610010099.510097.39999.910096.395.591.297.710098.4—
CFlow2021.1198.310097.697.798.799.610010099.396.891.999.999.795.299.198.5——
CFLOW-AD2022.1098.399.399.610099.999.110097.697.710099.396.891.999.795.298.5——
CS-FlowType=Embedding-based2023.0398.21009910010010099.899.197.199.699.198.697.691.999.399.798.799.8
DSR2022.0898.210010010010096.310093.898.195.698.597.596.299.797.8100——
PEFMCategory=Normalizing Flow Based2022.1098.1310096.5710099.4999.1910098.9591.999.8999.8597.5196.4396.3897.8398.03——
RevDistType=Embedding-based2023.039898.910010099.399.21009596.399.910096.69799.596.798.598.599.5
DRAEM2022.08989799.910099.699.199.291.898.510098.798.993.910093.1100——
DRAEM2023.0898—————————————————
PaDiM2021.1197.9—————————————————
FastFlowCategory=Normalizing Flow Based2022.1097.5299.1599.6810010099.5610097.5898.5297.9699.5198.2286.3489.1698.5898.55——
DRÆMType=Synthesizing-based2023.0397.49799.910099.699.199.291.898.510098.798.993.910093.11009899.1
CFLOW-ADCategory=Normalizing Flow Based2022.1097.2499.0496.2410010099.310097.4197.6910098.9296.9283.9592.7897.3899.03——
OmniALEvaluation Protocol=Unified case2024.0397.298.799.99999.693.210098.295.295.699.297.28810093.810097.2—
OmniAL2024.0397.298.799.99999.693.210098.295.295.699.297.28810093.8100——
Cut Paste2021.1197.110096.295.410099.199.910098.693.386.699.890.797.599.899.9——
CDO2023.0896.8—————————————————
UniADEvaluation Protocol=Unified case2024.0396.599.898.210099.398.699.795.286.999.899.293.787.594.299.895.896.5—
UniAD2024.0396.599.898.210099.398.699.795.286.999.899.293.787.594.299.895.8——
SPADE*2021.1196.298.69999.589.895.898.193.298.698.996.996.599.598.98198.8——
[11]2022.0896.193.910010094.699.198.281.298.298.399.994.388.799.496.199.9——
PaDiMType=Embedding-based2023.039699.896.710098.199.299.197.187.599.496.290.197.510094.498.695.895.5
DRATraining setting=ten random anomaly examples, Protocol=general setting2022.0395.99498.710099.499.810090.993.510099.790.497.782.691.5100——
DRAtraining_anomaly_examples=10, protocol=general setting2022.0395.99498.710099.499.810090.993.510099.798.482.682.691.5100——
CutPasteType=Synthesizing-based2023.0395.593.910010094.699.198.281.298.298.399.994.993.999.496.199.996.197.5
[7]2022.0895.599.995.710097.498.899.892.291.593.399.294.484.487.297.890.9——
Patchcore2024.0395.396.197.110099.998.410099.285.610099.993.382.988.996.791.9——
DifferNet2021.1194.98497.199.492.999.89986.988.899.195.195.999.396.196.398.6——
DifferNet2022.1094.992.98497.199.499.89995.986.999.396.188.896.398.691.195.1——
DevNettraining_anomaly_examples=10, protocol=general setting2022.0394.586.796.799.998.799.999.389.286.510099.186.6868692.499——
PMADEvaluation Protocol=Unified case2024.0394.59996.210099.899.699.893.580.599.69889.473.395.897.29694.5—
[17]2022.0894.495.592.910097.497.61009492.393.191.283.381.293.895.997.9——
EAR + KDScale=m*, Masking=Deterministic single saliency, Hint=✓, Knowledge Distillation=True (SQUID)2023.1094.289.995.910096.598.599.787.18799.787.692.288.610094.795.5——
MINNSTraining setting=ten random anomaly examples, Protocol=general setting2022.0393.987.698.399.39899.899.584.290.510098.491.393.28188.998.1——
FLOStraining_anomaly_examples=10, protocol=general setting2022.0393.97896.699.395.210099.58990.210098.492.9909086.299.5——
EARScale=m*, Masking=Deterministic single saliency, Hint=✓, Knowledge Distillation=/, Training (sec)=3109, Inference (msec)=1972023.1093.48897.610095.699.599.585.586.999.686.887.385.410093.396.1——
SAOEtraining_anomaly_examples=10, protocol=general setting2022.0392.675.595.210094.497.699.886.28510097.687.286.586.58696.5——
EARScale=m, Masking=Deterministic single saliency, Hint=✓, Knowledge Distillation=/, Training (sec)=3109, Inference (msec)=1972023.1092.284.697.610095.698.699.485.186.999.286.88777.610089.595.1——
EAR + KDScale=m, Masking=Deterministic single saliency, Hint=✓, Knowledge Distillation=True (SQUID)2023.1092.28595.210091.894.699.785.38799.785.692.277.910094.794.9——
PatchSVDD2021.1192.192.994.690.997.896.598.690.376.7929486.181.310091.597.9——
EARw/o attnMasking=/, Hint=✓, Knowledge Distillation=/, Training (sec)=3078, Inference (msec)=372023.109271.896.310093.994.510088.891.899.684.186.782.510094.396.3——
FastFlowEvaluation Protocol=Unified case2024.0391.891.685.793.799.29810090.990.598.996.590.476.886.185.793.891.8—
[22]2022.0891.784.299.610098.79399.981.988.483.388.583.884.510090.998.1——
RIADMasking=✓ (multi), Hint=/, Knowledge Distillation=/, Training (sec)=35478, Inference (msec)=3662023.1091.784.299.610098.79395.981.988.483.388.583.884.510090.998.1——
AE-SSIMType=Reconstruction-based2023.03918794785973938294978991969290888778
MLEPtraining_anomaly_examples=10, protocol=general setting2022.0390.778.19881.398.899.998.185.781.810097.484.578.378.392.798.4——
RIADType=Reconstruction-based2023.0389.984.299.610098.79399.981.988.483.388.583.884.510090.998.191.795.1
EARw/o obfMasking=Deterministic single saliency, Hint=/, Knowledge Distillation=/, Training (sec)=3084, Inference (msec)=1562023.1089.184.896.910092.890.499.579.578.498.683.273.88099.489.190——
CFLOWEvaluation Protocol=Unified case2024.038998.895.910097.99998.780.475.597.187.88859.57886.792.289—
DRAtraining_anomaly_examples=1, protocol=general setting2022.0388.385.997.298.996.598.510087.663.197.794.883.290.36566.899——
DRAEMEvaluation Protocol=Unified case2024.0388.19899.398.799.899.897.557.865.393.772.882.29290.674.898.888.1—
DRAEM2024.0388.19899.398.799.899.897.557.865.393.772.882.29290.674.898.8——
[4]2022.0887.791.68188.299.197.79986.286.193.18287.954.995.381.891.9——
KDADtraining_anomaly_examples=Unsupervised, protocol=general setting2022.0386.177.474.994.891.19499.28877.598.474.382.480.586.380.592.7——
SimpleNet2024.0385.687.679.195.297.997.586.571.577.894.387.880.272.887.879.788.5——
PaDiMEvaluation Protocol=Unified case2024.0384.293.873.999.993.398.497.970.973.485.58868.856.995.386.679.784.2—
PaDiM2024.0384.293.873.999.993.398.497.970.973.485.58868.856.995.386.679.7——
SAOEtraining_anomaly_examples=1, protocol=general setting2022.0383.476.692.199.693.594.898.981.961.196.192.265.265.368.66897——
DeepSADTraining setting=ten random anomaly examples, Protocol=general setting2022.038379.185.483.388.878.191.369.647.61008687.577.488.586.897.4——
MKDEvaluation Protocol=Unified case2024.0381.969.883.893.689.593.498.778.268.397.164.979.775.675.373.487.481.9—
MKD2024.0381.969.883.893.689.593.498.778.268.397.164.979.775.675.373.487.4——
DevNettraining_anomaly_examples=1, protocol=general setting2022.0379.474.689.187.375.29097.678.356.497.987.676.939.975.372.292.2——
FLOStraining_anomaly_examples=1, protocol=general setting2022.0379.275.587.179.178.792.797.57966.697.69374.533.773.170.988.5——
MLEPtraining_anomaly_examples=1, protocol=general setting2022.0374.470.183.978.192.76692.768.855.870.487.865.667.561.769.585.6——
PromptADmode=Few-shot2026.01—99.5497.9999.3496.0595.4497.3194.195.698.2191.6593.6395.1998.6385.6590.77——
Proposedmode=Few-normal shot, scoring=deviation-based patch-wise2026.01—99.5497.9799.3696.0695.2997.2893.7996.2598.2792.6795.2694.1198.6785.6692.04——
WinCLIPmode=Few-shot2026.01—99.196.399.295.794.896.893.696.0497.7689.593.895.99784.391.4——