Backbones

Overview

The backbones module provides access to a wide variety of pre-trained neural network architectures for feature extraction in PatchCore. These backbones are loaded with ImageNet pre-trained weights.

Functions

load

def load(name: str) -> torch.nn.Module

Loads a pre-trained backbone network by name.

str

required

Name of the backbone architecture to load. See Available Backbones for the complete list.

Returns

torch.nn.Module

Pre-trained PyTorch model loaded with ImageNet weights.

Example

import patchcore.backbones

# Load WideResNet50 backbone
backbone = patchcore.backbones.load("wideresnet50")

# Load ResNet50 backbone
backbone = patchcore.backbones.load("resnet50")

# Load Vision Transformer
backbone = patchcore.backbones.load("vit_base")

Available Backbones

The following pre-trained backbones are available:

ResNet Family

str

ResNet-50 architecture from torchvision

str

ResNet-101 architecture from torchvision

str

ResNet-200 architecture from timm

str

ResNeXt-101 (32x8d) architecture from torchvision

str

ResNeSt-50 (4s2x40d) architecture from timm

Wide ResNet Family

str

Wide ResNet-50-2 architecture from torchvision. Recommended for PatchCore.

str

Wide ResNet-101-2 architecture from torchvision

ResNetV2 (BiT) Family

str

ResNetV2-50x3 trained with Big Transfer (BiT) on ImageNet-1k

str

ResNetV2-50x3 trained with BiT on ImageNet-21k

str

ResNetV2-101x3 trained with BiT on ImageNet-1k

str

ResNetV2-101x3 trained with BiT on ImageNet-21k

str

ResNetV2-101 standard architecture

str

ResNetV2-152x4 trained with BiT on ImageNet-1k

str

ResNetV2-152x4 trained with BiT on ImageNet-21k

str

ResNetV2-152x2 teacher model with 384x384 input resolution

VGG Family

str

VGG-11 architecture from torchvision

str

VGG-19 architecture from torchvision

str

VGG-19 with batch normalization from torchvision

AlexNet

str

AlexNet architecture from torchvision

DenseNet Family

str

DenseNet-121 architecture from timm

str

DenseNet-201 architecture from timm

EfficientNet Family

str

EfficientNet-B1 (TensorFlow weights) from timm

str

EfficientNet-B3 (TensorFlow weights) from timm

str

EfficientNet-B3a architecture from timm

str

EfficientNet-B5 (TensorFlow weights) from timm

str

EfficientNet-B7 (TensorFlow weights) from timm

str

EfficientNetV2-M (TensorFlow weights) from timm

str

EfficientNetV2-L (TensorFlow weights) from timm

MNASNet Family

str

MNASNet 1.0 architecture from timm

str

MNASNet-A1 architecture from timm

str

MNASNet-B1 architecture from timm

Vision Transformer (ViT) Family

str

Vision Transformer Small with 16x16 patches (224x224 input)

str

Vision Transformer Base with 16x16 patches (224x224 input)

str

Vision Transformer Large with 16x16 patches (224x224 input)

str

Vision Transformer Large with ResNet-50 hybrid backbone

DeiT Family

str

Data-efficient Image Transformer (DeiT) Base with 16x16 patches

str

DeiT Base with distillation token (16x16 patches)

Swin Transformer Family

str

Swin Transformer Base with 4x4 patches and 7x7 window size

str

Swin Transformer Large with 4x4 patches and 7x7 window size

Inception

str

Inception-V4 architecture from timm

BN-Inception

str

Batch Normalized Inception trained on ImageNet (requires pretrainedmodels package)

Usage with PatchCore

import torch
import patchcore.backbones
from patchcore.patchcore import PatchCore

# Initialize device
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

# Load backbone
backbone = patchcore.backbones.load("wideresnet50")

# Create PatchCore model
model = PatchCore(device)
model.load(
    backbone=backbone,
    layers_to_extract_from=["layer2", "layer3"],
    device=device,
    input_shape=(3, 224, 224),
    pretrain_embed_dimension=1024,
    target_embed_dimension=1024
)

Recommended Backbone: The original PatchCore paper uses wideresnet50 with features extracted from layer2 and layer3, which provides excellent performance for most industrial anomaly detection tasks.

Backbone Registry

The complete mapping of backbone names to their implementations is stored in the _BACKBONES dictionary:

_BACKBONES = {
    "alexnet": "models.alexnet(pretrained=True)",
    "resnet50": "models.resnet50(pretrained=True)",
    "wideresnet50": "models.wide_resnet50_2(pretrained=True)",
    "vit_base": 'timm.create_model("vit_base_patch16_224", pretrained=True)',
    # ... and many more
}

Dependencies

torchvision: For ResNet, VGG, AlexNet, and Wide ResNet models
timm: For Vision Transformers, EfficientNets, and many other architectures
pretrainedmodels (optional): For BN-Inception

All backbones are loaded with pre-trained weights. The pretrained=True parameter ensures that ImageNet weights are used, which is crucial for transfer learning in anomaly detection.

Core Components

Utilities

CLI Tools

Overview

Functions

load

Returns

Example

Available Backbones

ResNet Family

Wide ResNet Family

ResNetV2 (BiT) Family

VGG Family

AlexNet

DenseNet Family

EfficientNet Family

MNASNet Family

Vision Transformer (ViT) Family

DeiT Family

Swin Transformer Family

Inception

BN-Inception

Usage with PatchCore

Backbone Registry

Dependencies

​Overview

​Functions

​load

​Returns

​Example

​Available Backbones

​ResNet Family

​Wide ResNet Family

​ResNetV2 (BiT) Family

​VGG Family

​AlexNet

​DenseNet Family

​EfficientNet Family

​MNASNet Family

​Vision Transformer (ViT) Family

​DeiT Family

​Swin Transformer Family

​Inception

​BN-Inception

​Usage with PatchCore

​Backbone Registry

​Dependencies

Overview

Functions

load

Returns

Example

Available Backbones

ResNet Family

Wide ResNet Family

ResNetV2 (BiT) Family

VGG Family

AlexNet

DenseNet Family

EfficientNet Family

MNASNet Family

Vision Transformer (ViT) Family

DeiT Family

Swin Transformer Family

Inception

BN-Inception

Usage with PatchCore

Backbone Registry

Dependencies