Skip to content
forked from TooTouch/MemSeg

Unofficial re-implementation of MemSeg for Anomaly Detection

License

Notifications You must be signed in to change notification settings

souxun2015/MemSeg

 
 

Repository files navigation

MemSeg

Unofficial re-implementation for MemSeg: A semi-supervised method for image surface defect detection using differences and commonalities

Environments

  • Docker image: nvcr.io/nvidia/pytorch:20.12-py3
einops==0.5.0
timm==0.5.4
wandb==0.12.17
omegaconf
imgaug==0.4.0

Process

1. Anomaly Simulation Strategy

2. Model Process

Run

Example

python main.py configs=configs.yaml DATASET.target=bottle

Demo

voila "[demo] model inference.ipynb" --port ${port} --Voila.ip ${ip}

Results

  • Backbone: ResNet18
target AUROC-image AUROC-pixel AUPRO-pixel
leather 100 98.31 99.05
pill 96.21 88 90.23
carpet 98.72 94.1 95.31
hazelnut 97.89 89.28 94.86
tile 100 98.97 98.84
cable 83.71 74.69 73.21
toothbrush 100 98.67 97.13
transistor 92.42 75 79.41
zipper 99.63 93.94 93
metal_nut 90.42 80.99 90.62
grid 99.92 96.48 95.87
bottle 100 94.67 92.61
capsule 92.34 83.45 84.34
screw 81.64 83.04 82.93
wood 99.74 94.9 94.45
Average 95.51 89.63 90.79

Citation

@article{DBLP:journals/corr/abs-2205-00908,
  author    = {Minghui Yang and
               Peng Wu and
               Jing Liu and
               Hui Feng},
  title     = {MemSeg: {A} semi-supervised method for image surface defect detection
               using differences and commonalities},
  journal   = {CoRR},
  volume    = {abs/2205.00908},
  year      = {2022},
  url       = {https://doi.org/10.48550/arXiv.2205.00908},
  doi       = {10.48550/arXiv.2205.00908},
  eprinttype = {arXiv},
  eprint    = {2205.00908},
  timestamp = {Tue, 03 May 2022 15:52:06 +0200},
  biburl    = {https://dblp.org/rec/journals/corr/abs-2205-00908.bib},
  bibsource = {dblp computer science bibliography, https://dblp.org}
}

About

Unofficial re-implementation of MemSeg for Anomaly Detection

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 99.3%
  • Other 0.7%