βNow has streamlit support!β Run $ streamlit run streamlit_app.py
This repo aims to reproduce the results of the following KNN-based anomaly detection methods:
- SPADE (Cohen et al. 2021) - knn in z-space and distance to feature maps
- PaDiM* (Defard et al. 2020) - distance to multivariate Gaussian of feature maps
- PatchCore (Roth et al. 2021) - knn distance to avgpooled feature maps
* actually does not have any knn mechanism, but shares many things implementation-wise.
$ pipenv install -r requirements.txt
CLI:
$ python indad/run.py METHOD [--dataset DATASET]
Results can be found under ./results/
.
Code example:
from indad.model import SPADE
model = SPADE(k=5, backbone_name="resnet18")
# feed healthy dataset
model.fit(...)
# get predictions
img_lvl_anom_score, pxl_lvl_anom_score = model.predict(...)
ποΈ
Check out one of the downloaded MVTec datasets. Naming of images should correspond among folders. Right now there is no support for no ground truth pixel masks.
πdatasets
β πyour_custom_dataset
β£ π ground_truth/defective
β β£ π defect_type_1
β β π defect_type_2
β£ π test
β β£ π defect_type_1
β β£ π defect_type_2
β β π good
β π train/good
$ python indad/run.py METHOD --dataset your_custom_dataset
π = paper, π = this repo
class | SPADE π | SPADE π | PaDiM π | PaDiM π | PatchCore π | PatchCore π |
---|---|---|---|---|---|---|
bottle | - | 98.8 | 99.8 | β 100.0β | β 100.0β | |
cable | - | 76.5 | 93.3 | β 99.5β | 96.2 | |
capsule | - | 84.6 | 88.3 | 98.1 | 95.3 | |
carpet | - | 84.3 | β 99.4 | 98.7 | 98.7 | |
grid | - | 37.1 | 98.2 | β 98.2β | 93.0 | |
hazelnut | - | 88.7 | 83.7 | β 100.0β | 100.0 | |
leather | - | 97.1 | 99.9 | β 100.0β | 100.0 | |
metal_nut | - | 74.6 | 99.4 | β 100.0β | 98.3 | |
pill | - | 72.6 | 89.0 | β 96.6β | 92.8 | |
screw | - | 53.1 | 83.0 | 98.1 | 96.7 | |
tile | - | 97.8 | 98.6 | 98.7 | β 99.0β | |
toothbrush | - | 89.4 | 97.2 | β 100.0β | 98.1 | |
transistor | - | 89.2 | 96.8 | β 100.0β | 99.7 | |
wood | - | 98.3 | 98.9 | β 99.2β | 98.8 | |
zipper | - | 96.7 | 89.5 | β 99.4β | 98.4 | |
averages | 85.5 | 82.6 | 95.3* | 94.3 | β 99.1β | 97.7 |
- PaDiM average referencing PaDiM-WR50-Rd550
class | SPADE π | SPADE π | PaDiM π | PaDiM π | PatchCore π | PatchCore π |
---|---|---|---|---|---|---|
bottle | 97.5 | 97.7 | 98.3 | 97.8 | β 98.6β | 97.8 |
cable | 93.7 | 94.3 | 96.7 | 96.1 | β 98.5β | 97.4 |
capsule | 97.6 | 98.6 | 98.5 | 98.3 | β 98.9β | 98.3 |
carpet | 87.4 | 99.0 | 99.1 | 98.6 | β 99.1β | 98.3 |
grid | 88.5 | 96.1 | 97.3 | 97.2 | β 98.7β | 96.7 |
hazelnut | 98.4 | 98.1 | 98.2 | 97.5 | β 98.7β | 98.1 |
leather | 97.2 | 99.2 | 99.2 | 98.7 | β 99.3β | 98.4 |
metal_nut | β 99.0β | 96.1 | 97.2 | 96.5 | 98.4 | 96.2 |
pill | β 99.1β | 93.5 | 95.7 | 93.2 | 97.6 | 98.7 |
screw | 98.1 | 98.9 | 98.5 | 97.8 | β 99.4β | 98.4 |
tile | β 96.5β | 93.3 | 94.1 | 94.8 | 95.9 | 94.0 |
toothbrush | β 98.9β | β 98.9β | 98.8 | 98.3 | 98.7 | 98.1 |
transistor | β 97.9β | 96.3 | 97.5 | 97.2 | 96.4 | 97.5 |
wood | 94.1 | 94.4 | 94.7 | 93.6 | β 95.1β | 91.9 |
zipper | 96.5 | 98.2 | 98.5 | 97.4 | β 98.9β | 97.6 |
averages | 96.9 | 96.8 | 97.5 | 96.9 | β 98.1β | 97.2 |
PatchCore-10 was used.
The following parameters were used to calculate the results. They more or less correspond to the parameters used in the papers.
spade:
backbone: wide_resnet50_2
k: 50
padim:
backbone: wide_resnet50_2
d_reduced: 250
epsilon: 0.04
patchcore:
backbone: wide_resnet50_2
f_coreset: 0.1
n_reweight: 3
- Datasets
- Code skeleton
- Config files
- CLI
- Logging
- SPADE
- PADIM
- PatchCore
- Add custom dataset option
- Add dataset progress bar
- Add schematics
- Unit tests
- Data is processed in single images to avoid batch statistics interference.
- I decided to implement greedy kcenter from scratch and there is room for improvement.
torch.nn.AdaptiveAvgPool2d
for feature map resizing,torch.nn.functional.interpolate
for score map resizing.- GPU is used for backbones and coreset selection. GPU coreset selection currently runs at:
- 400-500 it/s @ float32 (RTX3080)
- 1000+ it/s @ float16 (RTX3080)
- hcw-00 for tipping
sklearn.random_projection.SparseRandomProjection
. - h1day for adding a custom range to the streamlit app.
SPADE:
@misc{cohen2021subimage,
title={Sub-Image Anomaly Detection with Deep Pyramid Correspondences},
author={Niv Cohen and Yedid Hoshen},
year={2021},
eprint={2005.02357},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
PaDiM:
@misc{defard2020padim,
title={PaDiM: a Patch Distribution Modeling Framework for Anomaly Detection and Localization},
author={Thomas Defard and Aleksandr Setkov and Angelique Loesch and Romaric Audigier},
year={2020},
eprint={2011.08785},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
PatchCore:
@misc{roth2021total,
title={Towards Total Recall in Industrial Anomaly Detection},
author={Karsten Roth and Latha Pemula and Joaquin Zepeda and Bernhard SchΓΆlkopf and Thomas Brox and Peter Gehler},
year={2021},
eprint={2106.08265},
archivePrefix={arXiv},
primaryClass={cs.CV}
}