Skip to content

An anomaly detection library comprising state-of-the-art algorithms and features such as experiment management, hyper-parameter optimization, and edge inference.

License

Notifications You must be signed in to change notification settings

openvinotoolkit/anomalib

Repository files navigation

A library for benchmarking, developing and deploying deep learning anomaly detection algorithms


Key Features • Getting Started • Docs • License

python pytorch openvino black Nightly-regression Test Pre-merge Checks Build Docs


Introduction

Anomalib is a deep learning library that aims to collect state-of-the-art anomaly detection algorithms for benchmarking on both public and private datasets. Anomalib provides several ready-to-use implementations of anomaly detection algorithms described in the recent literature, as well as a set of tools that facilitate the development and implementation of custom models. The library has a strong focus on image-based anomaly detection, where the goal of the algorithm is to identify anomalous images, or anomalous pixel regions within images in a dataset. Anomalib is constantly updated with new algorithms and training/inference extensions, so keep checking!

Sample Image

Key features:

  • The largest public collection of ready-to-use deep learning anomaly detection algorithms and benchmark datasets.
  • PyTorch Lightning based model implementations to reduce boilerplate code and limit the implementation efforts to the bare essentials.
  • All models can be exported to OpenVINO Intermediate Representation (IR) for accelerated inference on intel hardware.
  • A set of inference tools for quick and easy deployment of the standard or custom anomaly detection models.

Getting Started

To get an overview of all the devices where anomalib as been tested thoroughly, look at the Supported Hardware section in the documentation.

PyPI Install

You can get started with anomalib by just using pip.

pip install anomalib

Local Install

It is highly recommended to use virtual environment when installing anomalib. For instance, with anaconda, anomalib could be installed as,

yes | conda create -n anomalib_env python=3.8
conda activate anomalib_env
git clone https://github.com/openvinotoolkit/anomalib.git
cd anomalib
pip install -e .

Training

By default python tools/train.py runs PADIM model on leather category from the MVTec AD (CC BY-NC-SA 4.0) dataset.

python tools/train.py    # Train PADIM on MVTec AD leather

Training a model on a specific dataset and category requires further configuration. Each model has its own configuration file, config.yaml , which contains data, model and training configurable parameters. To train a specific model on a specific dataset and category, the config file is to be provided:

python tools/train.py --config <path/to/model/config.yaml>

For example, to train PADIM you can use

python tools/train.py --config anomalib/models/padim/config.yaml

Note that --model_config_path will be deprecated in v0.2.8 and removed in v0.2.9.

Alternatively, a model name could also be provided as an argument, where the scripts automatically finds the corresponding config file.

python tools/train.py --model padim

where the currently available models are:

Custom Dataset

It is also possible to train on a custom folder dataset. To do so, data section in config.yaml is to be modified as follows:

dataset:
  name: <name-of-the-dataset>
  format: folder
  path: <path/to/folder/dataset>
  normal: normal # name of the folder containing normal images.
  abnormal: abnormal # name of the folder containing abnormal images.
  task: segmentation # classification or segmentation
  mask: <path/to/mask/annotations> #optional
  extensions: null
  split_ratio: 0.2  # ratio of the normal images that will be used to create a test split
  seed: 0
  image_size: 256
  train_batch_size: 32
  test_batch_size: 32
  num_workers: 8
  transform_config: null
  create_validation_set: true
  tiling:
    apply: false
    tile_size: null
    stride: null
    remove_border_count: 0
    use_random_tiling: False
    random_tile_count: 16

Inference

Anomalib contains several tools that can be used to perform inference with a trained model. The script in tools/inference contains an example of how the inference tools can be used to generate a prediction for an input image.

If the specified weight path points to a PyTorch Lightning checkpoint file (.ckpt), inference will run in PyTorch. If the path points to an ONNX graph (.onnx) or OpenVINO IR (.bin or .xml), inference will run in OpenVINO.

The following command can be used to run inference from the command line:

python tools/inference.py \
    --config <path/to/model/config.yaml> \
    --weight_path <path/to/weight/file> \
    --image_path <path/to/image>

As a quick example:

python tools/inference.py \
    --config anomalib/models/padim/config.yaml \
    --weight_path results/padim/mvtec/bottle/weights/model.ckpt \
    --image_path datasets/MVTec/bottle/test/broken_large/000.png

If you want to run OpenVINO model, ensure that openvino apply is set to True in the respective model config.yaml.

optimization:
  openvino:
    apply: true

Example OpenVINO Inference:

python tools/inference.py \
    --config  \
    anomalib/models/padim/config.yaml  \
    --weight_path  \
    results/padim/mvtec/bottle/compressed/compressed_model.xml  \
    --image_path  \
    datasets/MVTec/bottle/test/broken_large/000.png  \
    --meta_data  \
    results/padim/mvtec/bottle/compressed/meta_data.json

Ensure that you provide path to meta_data.json if you want the normalization to be applied correctly.


Datasets

anomalib supports MVTec AD (CC BY-NC-SA 4.0) and BeanTech (CC-BY-SA) for benchmarking and folder for custom dataset training/inference.

MVTec AD dataset is one of the main benchmarks for anomaly detection, and is released under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (CC BY-NC-SA 4.0).

Image-Level AUC

Model Avg Carpet Grid Leather Tile Wood Bottle Cable Capsule Hazelnut Metal Nut Pill Screw Toothbrush Transistor Zipper
PatchCore Wide ResNet-50 0.980 0.984 0.959 1.000 1.000 0.989 1.000 0.990 0.982 1.000 0.994 0.924 0.960 0.933 1.000 0.982
PatchCore ResNet-18 0.973 0.970 0.947 1.000 0.997 0.997 1.000 0.986 0.965 1.000 0.991 0.916 0.943 0.931 0.996 0.953
CFlow Wide ResNet-50 0.962 0.986 0.962 1.0 0.999 0.993 1.0 0.893 0.945 1.0 0.995 0.924 0.908 0.897 0.943 0.984
PaDiM Wide ResNet-50 0.950 0.995 0.942 1.0 0.974 0.993 0.999 0.878 0.927 0.964 0.989 0.939 0.845 0.942 0.976 0.882
PaDiM ResNet-18 0.891 0.945 0.857 0.982 0.950 0.976 0.994 0.844 0.901 0.750 0.961 0.863 0.759 0.889 0.920 0.780
STFPM Wide ResNet-50 0.876 0.957 0.977 0.981 0.976 0.939 0.987 0.878 0.732 0.995 0.973 0.652 0.825 0.5 0.875 0.899
STFPM ResNet-18 0.893 0.954 0.982 0.989 0.949 0.961 0.979 0.838 0.759 0.999 0.956 0.705 0.835 0.997 0.853 0.645
DFM Wide ResNet-50 0.891 0.978 0.540 0.979 0.977 0.974 0.990 0.891 0.931 0.947 0.839 0.809 0.700 0.911 0.915 0.981
DFM ResNet-18 0.894 0.864 0.558 0.945 0.984 0.946 0.994 0.913 0.871 0.979 0.941 0.838 0.761 0.95 0.911 0.949
DFKDE Wide ResNet-50 0.774 0.708 0.422 0.905 0.959 0.903 0.936 0.746 0.853 0.736 0.687 0.749 0.574 0.697 0.843 0.892
DFKDE ResNet-18 0.762 0.646 0.577 0.669 0.965 0.863 0.951 0.751 0.698 0.806 0.729 0.607 0.694 0.767 0.839 0.866
GANomaly 0.421 0.203 0.404 0.413 0.408 0.744 0.251 0.457 0.682 0.537 0.270 0.472 0.231 0.372 0.440 0.434

Pixel-Level AUC

Model Avg Carpet Grid Leather Tile Wood Bottle Cable Capsule Hazelnut Metal Nut Pill Screw Toothbrush Transistor Zipper
PatchCore Wide ResNet-50 0.980 0.988 0.968 0.991 0.961 0.934 0.984 0.988 0.988 0.987 0.989 0.980 0.989 0.988 0.981 0.983
PatchCore ResNet-18 0.976 0.986 0.955 0.990 0.943 0.933 0.981 0.984 0.986 0.986 0.986 0.974 0.991 0.988 0.974 0.983
CFlow Wide ResNet-50 0.971 0.986 0.968 0.993 0.968 0.924 0.981 0.955 0.988 0.990 0.982 0.983 0.979 0.985 0.897 0.980
PaDiM Wide ResNet-50 0.979 0.991 0.970 0.993 0.955 0.957 0.985 0.970 0.988 0.985 0.982 0.966 0.988 0.991 0.976 0.986
PaDiM ResNet-18 0.968 0.984 0.918 0.994 0.934 0.947 0.983 0.965 0.984 0.978 0.970 0.957 0.978 0.988 0.968 0.979
STFPM Wide ResNet-50 0.903 0.987 0.989 0.980 0.966 0.956 0.966 0.913 0.956 0.974 0.961 0.946 0.988 0.178 0.807 0.980
STFPM ResNet-18 0.951 0.986 0.988 0.991 0.946 0.949 0.971 0.898 0.962 0.981 0.942 0.878 0.983 0.983 0.838 0.972

Image F1 Score

Model Avg Carpet Grid Leather Tile Wood Bottle Cable Capsule Hazelnut Metal Nut Pill Screw Toothbrush Transistor Zipper
PatchCore Wide ResNet-50 0.976 0.971 0.974 1.000 1.000 0.967 1.000 0.968 0.982 1.000 0.984 0.940 0.943 0.938 1.000 0.979
PatchCore ResNet-18 0.970 0.949 0.946 1.000 0.98 0.992 1.000 0.978 0.969 1.000 0.989 0.940 0.932 0.935 0.974 0.967
CFlow Wide ResNet-50 0.944 0.972 0.932 1.0 0.988 0.967 1.0 0.832 0.939 1.0 0.979 0.924 0.971 0.870 0.818 0.967
PaDiM Wide ResNet-50 0.951 0.989 0.930 1.0 0.960 0.983 0.992 0.856 0.982 0.937 0.978 0.946 0.895 0.952 0.914 0.947
PaDiM ResNet-18 0.916 0.930 0.893 0.984 0.934 0.952 0.976 0.858 0.960 0.836 0.974 0.932 0.879 0.923 0.796 0.915
STFPM Wide ResNet-50 0.926 0.973 0.973 0.974 0.965 0.929 0.976 0.853 0.920 0.972 0.974 0.922 0.884 0.833 0.815 0.931
STFPM ResNet-18 0.932 0.961 0.982 0.989 0.930 0.951 0.984 0.819 0.918 0.993 0.973 0.918 0.887 0.984 0.790 0.908
DFM Wide ResNet-50 0.918 0.960 0.844 0.990 0.970 0.959 0.976 0.848 0.944 0.913 0.912 0.919 0.859 0.893 0.815 0.961
DFM ResNet-18 0.919 0.895 0.844 0.926 0.971 0.948 0.977 0.874 0.935 0.957 0.958 0.921 0.874 0.933 0.833 0.943
DFKDE Wide ResNet-50 0.875 0.907 0.844 0.905 0.945 0.914 0.946 0.790 0.914 0.817 0.894 0.922 0.855 0.845 0.722 0.910
DFKDE ResNet-18 0.872 0.864 0.844 0.854 0.960 0.898 0.942 0.793 0.908 0.827 0.894 0.916 0.859 0.853 0.756 0.916
GANomaly 0.834 0.864 0.844 0.852 0.836 0.863 0.863 0.760 0.905 0.777 0.894 0.916 0.853 0.833 0.571 0.881

Reference

If you use this library and love it, use this to cite it 🤗

@misc{anomalib,
      title={Anomalib: A Deep Learning Library for Anomaly Detection},
      author={Samet Akcay and
              Dick Ameln and
              Ashwin Vaidya and
              Barath Lakshmanan and
              Nilesh Ahuja and
              Utku Genc},
      year={2022},
      eprint={2202.08341},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}