Contamination MapVision / Model
WideResNet-50-2 (supervised ImageNet-1k, torchvision)
- Organisation
- PyTorch / torchvision
- Family
- ResNet · version IMAGENET1K_V1
- Released
- Unknown
- Checkpoints
- torchvision wide_resnet50_2 IMAGENET1K_V1 (wide_resnet50_2-95faca4d.pth)
- Weights
- Open · BSD-3-Clause (torchvision code)
- Links
- Code
Training data disclosure: Full list
Level determined for 100% of benchmark cells.
- TestTest data seen0
- SeenSeen, split unclear0
- TrainTraining data seen0
- Related data seen0
- NoNot seen8
- ?Unknown0
“@handle_legacy_interface(weights=("pretrained", Wide_ResNet50_2_Weights.IMAGENET1K_V1))”
Other declared training data
-
ImageNet-1K · split: unknown
Supervised classification weights. The split is not stated in these sources.
Evidence
“class Wide_ResNet50_2_Weights(WeightsEnum): IMAGENET1K_V1 = Weights(”
raw.githubusercontent.com · code · torchvision/models/resnet.py · retrieved 2026-10-15 “PatchCore uses a network φ pre-trained on ImageNet.”
arxiv.org · paper · Section 3.1 · retrieved 2026-10-15
Benchmark cells
Industrial anomaly detection
Not seen MVTec AD · absent from the complete list of training data
Evidence (1)
“@handle_legacy_interface(weights=("pretrained", Wide_ResNet50_2_Weights.IMAGENET1K_V1))”
Not seen MVTec AD 2 · absent from the complete list of training data
Evidence (1)
“@handle_legacy_interface(weights=("pretrained", Wide_ResNet50_2_Weights.IMAGENET1K_V1))”
Not seen MVTec LOCO AD · absent from the complete list of training data
Evidence (1)
“@handle_legacy_interface(weights=("pretrained", Wide_ResNet50_2_Weights.IMAGENET1K_V1))”
Not seen VisA · absent from the complete list of training data
Evidence (1)
“@handle_legacy_interface(weights=("pretrained", Wide_ResNet50_2_Weights.IMAGENET1K_V1))”
Not seen Real-IAD · absent from the complete list of training data
Evidence (1)
“@handle_legacy_interface(weights=("pretrained", Wide_ResNet50_2_Weights.IMAGENET1K_V1))”
Not seen MPDD · absent from the complete list of training data
Evidence (1)
“@handle_legacy_interface(weights=("pretrained", Wide_ResNet50_2_Weights.IMAGENET1K_V1))”
Not seen BTAD · absent from the complete list of training data
Evidence (1)
“@handle_legacy_interface(weights=("pretrained", Wide_ResNet50_2_Weights.IMAGENET1K_V1))”
Not seen KolektorSDD2 · absent from the complete list of training data
Evidence (1)
“@handle_legacy_interface(weights=("pretrained", Wide_ResNet50_2_Weights.IMAGENET1K_V1))”
Notes
PatchCore's code loads it as models.wide_resnet50_2(pretrained=True), which maps to IMAGENET1K_V1 (see backbones.py). The PatchCore paper says it 'uses a WideResNet50-backbone [57] for direct comparability'.
Sources
- torchvision resnet.py · code · retrieved 2026-10-15
- Towards Total Recall in Industrial Anomaly Detection (PatchCore) · paper · retrieved 2026-10-15
- patchcore-inspection backbones.py · code · retrieved 2026-10-15