Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/pangoraw/padim
:dolphin: Re-implementation of PaDiM and code for the article "Weakly Supervised Detection of Marine Animals in High Resolution Aerial Images"
https://github.com/pangoraw/padim
anomaly-detection mvtec-ad pytorch unsupervised-learning
Last synced: 26 days ago
JSON representation
:dolphin: Re-implementation of PaDiM and code for the article "Weakly Supervised Detection of Marine Animals in High Resolution Aerial Images"
- Host: GitHub
- URL: https://github.com/pangoraw/padim
- Owner: Pangoraw
- License: mit
- Created: 2021-03-03T13:37:50.000Z (over 3 years ago)
- Default Branch: main
- Last Pushed: 2023-01-27T15:55:04.000Z (almost 2 years ago)
- Last Synced: 2023-02-27T00:32:39.972Z (over 1 year ago)
- Topics: anomaly-detection, mvtec-ad, pytorch, unsupervised-learning
- Language: Python
- Homepage: https://arxiv.org/abs/2011.08785
- Size: 108 KB
- Stars: 19
- Watchers: 4
- Forks: 2
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# PaDiM
###### _A Patch Distribution Modeling Framework for Anomaly Detection and Localization_
This is an **unofficial** re-implementation of the paper *PaDiM: a Patch Distribution Modeling Framework for Anomaly Detection and Localization* available on [arxiv](http://arxiv.org/abs/2011.08785).
### Features
The key features of this implementation are:
- Constant memory footprint - training on more images does not result in more memory required
- Resumable learning - the training step can be stopped and then resumed with inference in-between
- Limited dependencies - apart from PyTorch, Torchvision and Numpy
### Variants
This repository also contains variants on the original PaDiM model:
- [PaDiMSVDD](https://github.com/Pangoraw/PaDiM/blob/main/padim/padim_svdd.py) uses a [Deep-SVDD](http://proceedings.mlr.press/v80/ruff18a.html) model instead of a multi-variate Gaussian distribution for the normal patch representation.
- [PaDiMShared](https://github.com/Pangoraw/PaDiM/blob/main/padim/padim_shared.py) shares the multi-variate Gaussian distribution between all patches instead of learning it only for specific coordinates.
### Installation
```
git clone https://github.com/Pangoraw/PaDiM.git padim
```
### Getting started
#### Training
```python
from torch.utils.data import DataLoader
from padim import PaDiM
# i) Initialize
padim = PaDiM(num_embeddings=100, device="cpu", backbone="resnet18")
# ii) Create a dataloader producing image tensors
dataloader = DataLoader(...)
# iii) Consume the data to learn the normal distribution
# Use PaDiM.train(...)
padim.train(dataloader)
# Or PaDiM.train_one_batch(...)
for imgs in dataloader:
padim.train_one_batch(imgs)
```
#### Testing
With the same `PaDiM` instance as in the [Training](#training) section:
```python
for new_imgs in test_dataloader:
distances = padim.predict(new_imgs)
# distances is a (n * c) matrix of the mahalanobis distances
# Compute metrics...
```
### Acknowledgements
This implementation was built on the work of:
- [The original *PaDiM* paper](http://arxiv.org/abs/2011.08785)
- [taikiinoue45/PaDiM](https://github.com/taikiinoue45/PaDiM)'s implementation - see section [Features](#features) for the main differences.