You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.
gaoyang Tang 287ec30c3e remove GPU operations 7 months ago
READ_mindspore remove GPU operations 7 months ago
docs commit v1.1, add patchcore and riad 8 months ago
examples remove GPU operations 7 months ago
.gitignore Initial commit 10 months ago
LICENSE Initial commit 10 months ago
README.md commit v1.1, add patchcore and riad 8 months ago
requirements.txt remove GPU operations 7 months ago
setup.py commit setup, version and docs. 10 months ago

README.md


READ (Reconstruction or Embedding based Anomaly Detection)

This repo is the mindspore version of READ, plz jump to https://git.openi.org.cn/OpenI/READ_pytorch for the pytorch version.

READ is an open source toolbox focused on unsupervised anomaly detection/localization tasks. By only training on the defect-free samples, READ is able to recognize defect samples or even localize anomalies on defect samples.

The purpose of this repo is to promote the research and application of unsupervised anomaly detection and localization algorithms. READ is designed to provide:

  • A unified interface for encapsulating diverse anomaly localization algorithms
  • High quality implementations of novel anomaly localization algorithms
  • Templates for using these algorithms in a detailed task

In addition, READ provides the benchmarks for validating novel unsupervised anomaly detection and localization algorithms for MVTec AD dataset.

Changelog

  • [Jan 25 2022] READ_mindspore v0.1.1 is Released!
    Please refer to ChangeLog for details and release history.
  • [Nov 07 2021] READ_mindspore v0.1.0 is Released!
    Please refer to ChangeLog for details and release history.

Installation

Install the latest version from the master branch on OpenI

pip install -U git+https://git.openi.org.cn/OpenI/READ_mindspore

Please follow the Installation document to get a detailed instruction.

Getting Started

Please follow the Getting Started document to run the provided demo tasks.

Localization examples (based on READ)

Supported Algorithms

Results

Implementation results on MVTec

  • Image-level anomaly detection accuracy (ROCAUC)
MVTec SPADE-R50 PaDiM-R50 STPM-R18 RIAD-R18 PatchCore-R18
Carpet 0.908 0.998 0.970 0.622 0.852
Grid 0.783 0.927 0.944 0.938 0.546
Leather 0.980 1.0 0.999 0.990 0.945
Tile 0.939 0.986 0.981 0.657 0.891
Wood 0.985 0.998 0.990 0.899 0.984
All texture classes 0.919 0.982 0.977 0.821 0.844
Bottle 0.999 1.0 1.0 0.985 1.0
Cable 0.795 0.978 0.921 0.719 0.923
Capsule 0.897 0.902 0.777 0.648 0.919
Hazelnut 1.0 0.945 0.998 0.746 0.999
Metal nut 0.980 0.987 0.998 0.806 0.983
Pill 0.855 0.947 0.942 0.732 0.953
Screw 0.929 0.844 0.774 0.452 0.885
Toothbrush 0.958 0.964 0.922 0.467 1.0
Transistor 0.868 0.993 0.907 0.847 0.989
Zipper 0.870 0.859 0.949 0.957 0.930
All object classes 0.915 0.942 0.919 0.736 0.997
All classes 0.916 0.955 0.938 0.764 0.946
  • Pixel-level anomaly detection accuracy (ROCAUC)
MVTec SPADE-R50 PaDiM-R50 STPM-R18 RIAD-R18 Patchcore-R18
Carpet 0.985 0.982 0.988 0.910 0.972
Grid 0.976 0.958 0.975 0.978 0.892
Leather 0.992 0.987 0.992 0.983 0.984
Tile 0.938 0.936 0.962 0.744 0.914
Wood 0.955 0.942 0.957 0.797 0.908
All texture classes 0.969 0.961 0.975 0.882 0.934
Bottle 0.975 0.980 0.983 0.976 0.971
Cable 0.944 0.980 0.944 0.840 0.973
Capsule 0.988 0.984 0.966 0.876 0.977
Hazelnut 0.992 0.981 0.988 0.928 0.981
Metal nut 0.979 0.965 0.946 0.767 0.965
Pill 0.954 0.973 0.974 0.833 0.980
Screw 0.993 0.984 0.975 0.767 0.985
Toothbrush 0.991 0.987 0.988 0.853 0.983
Transistor 0.839 0.981 0.808 0.973 0.973
Zipper 0.978 0.978 0.984 0.958 0.973
All object classes 0.963 0.979 0.956 0.877 0.976
All classes 0.965 0.973 0.962 0.879 0.962

License

This project is released under the Open-Intelligence Open Source License V1.1.

Contact

Please contact me if there is any question (Chao Zhang chao.zhang46@tcl.com).

About

Machine Vision Group, TCL Corporate Research(HK) Co., Ltd is the main developer of READ.

Any contributions to READ is welcome!

简介

基于TCL在智能制造上缺陷检测的成功经验,TCL集团工业研究院开源了第一个工业视觉无监督异常检测框架,具有算法丰富、开箱即用、精度保证等特点。

Python other

贡献者 (2)