• 设为首页
  • 点击收藏
  • 手机版
    手机扫一扫访问
    迪恩网络手机版
  • 关注官方公众号
    微信扫一扫关注
    公众号

gudovskiy/cflow-ad: Official PyTorch code for WACV 2022 paper "CFLOW-AD: Re ...

原作者: [db:作者] 来自: 网络 收藏 邀请

开源软件名称(OpenSource Name):

gudovskiy/cflow-ad

开源软件地址(OpenSource Url):

https://github.com/gudovskiy/cflow-ad

开源编程语言(OpenSource Language):

Python 100.0%

开源软件介绍(OpenSource Introduction):

PWC

CFLOW-AD: Real-Time Unsupervised Anomaly Detection with Localization via Conditional Normalizing Flows

WACV 2022 preprint:https://arxiv.org/abs/2107.12571

Abstract

Unsupervised anomaly detection with localization has many practical applications when labeling is infeasible and, moreover, when anomaly examples are completely missing in the train data. While recently proposed models for such data setup achieve high accuracy metrics, their complexity is a limiting factor for real-time processing. In this paper, we propose a real-time model and analytically derive its relationship to prior methods. Our CFLOW-AD model is based on a conditional normalizing flow framework adopted for anomaly detection with localization. In particular, CFLOW-AD consists of a discriminatively pretrained encoder followed by a multi-scale generative decoders where the latter explicitly estimate likelihood of the encoded features. Our approach results in a computationally and memory-efficient model: CFLOW-AD is faster and smaller by a factor of 10x than prior state-of-the-art with the same input setting. Our experiments on the MVTec dataset show that CFLOW-AD outperforms previous methods by 0.36% AUROC in detection task, by 1.12% AUROC and 2.5% AUPRO in localization task, respectively. We open-source our code with fully reproducible experiments.

BibTex Citation

If you like our paper or code, please cite its WACV 2022 preprint using the following BibTex:

@inproceedings{Gudovskiy_2022_WACV,
    author    = {Gudovskiy, Denis and Ishizaka, Shun and Kozuka, Kazuki},
    title     = {{CFLOW-AD}: Real-Time Unsupervised Anomaly Detection With Localization via Conditional Normalizing Flows},
    booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
    month     = {January},
    year      = {2022},
    pages     = {98-107}
}

Installation

Install all packages with this command:

$ python3 -m pip install -U -r requirements.txt

Datasets

We support MVTec AD dataset for anomaly localization in factory setting and Shanghai Tech Campus (STC) dataset with surveillance camera videos. Please, download dataset from URLs and extract to data folder or make symlink to that folder or change default data path in main.py).

Code Organization

  • ./custom_datasets - contains dataloaders for MVTec and STC
  • ./custom_models - contains pretrained feature extractors

Training Models

  • Run code by selecting class name, feature extractor, input size, flow model etc.
  • The commands below should reproduce our reference MVTec results using WideResnet-50 extractor:
python3 main.py --gpu 0 --pro -inp 512 --dataset mvtec --class-name bottle
python3 main.py --gpu 0 --pro -inp 256 --dataset mvtec --class-name cable
python3 main.py --gpu 0 --pro -inp 256 --dataset mvtec --class-name capsule
python3 main.py --gpu 0 --pro -inp 512 --dataset mvtec --class-name carpet
python3 main.py --gpu 0 --pro -inp 512 --dataset mvtec --class-name grid
python3 main.py --gpu 0 --pro -inp 256 --dataset mvtec --class-name hazelnut
python3 main.py --gpu 0 --pro -inp 512 --dataset mvtec --class-name leather
python3 main.py --gpu 0 --pro -inp 256 --dataset mvtec --class-name metal_nut
python3 main.py --gpu 0 --pro -inp 256 --dataset mvtec --class-name pill
python3 main.py --gpu 0 --pro -inp 512 --dataset mvtec --class-name screw
python3 main.py --gpu 0 --pro -inp 512 --dataset mvtec --class-name tile
python3 main.py --gpu 0 --pro -inp 512 --dataset mvtec --class-name toothbrush
python3 main.py --gpu 0 --pro -inp 128 --dataset mvtec --class-name transistor
python3 main.py --gpu 0 --pro -inp 512 --dataset mvtec --class-name wood
python3 main.py --gpu 0 --pro -inp 512 --dataset mvtec --class-name zipper

Testing Pretrained Models

  • Download pretrained weights from Google Drive
  • The command below should reproduce MVTec results using light-weight MobileNetV3L extractor (AUROC, AUPRO) = (98.38%, 94.72%):
python3 main.py --gpu 0 --pro -enc mobilenet_v3_large --dataset mvtec --action-type norm-test -inp INPUT --class-name CLASS --checkpoint PATH/FILE.PT

CFLOW-AD Architecture

CFLOW-AD

Reference CFLOW-AD Results for MVTec

CFLOW-AD




鲜花

握手

雷人

路过

鸡蛋
该文章已有0人参与评论

请发表评论

全部评论

专题导读
热门推荐
阅读排行榜

扫描微信二维码

查看手机版网站

随时了解更新最新资讯

139-2527-9053

在线客服(服务时间 9:00~18:00)

在线QQ客服
地址:深圳市南山区西丽大学城创智工业园
电邮:jeky_zhao#qq.com
移动电话:139-2527-9053

Powered by 互联科技 X3.4© 2001-2213 极客世界.|Sitemap