Puzzle-AE: Novelty Detection In Images Through Solving Puzzles

This repository contains code for training and evaluating the proposed method in our paper Puzzle-AE: Novelty Detection in Images through Solving Puzzles.

Citation

If you find this useful for your research, please cite the following paper:

@misc{salehi2020puzzleae,
      title={Puzzle-AE: Novelty Detection in Images through Solving Puzzles}, 
      author={Mohammadreza Salehi and Ainaz Eftekhar and Niousha Sadjadi and Mohammad Hossein Rohban and Hamid R. Rabiee},
      year={2020},
      eprint={2008.12959},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

Clone this repo:

git clone https://github.com/Niousha12/Puzzle_Anomaly_Detection.git
cd Puzzle_Anomaly_Detection

Datsets:

This repository performs Novelty/Anomaly Detection in the following datasets: MNIST, Fashion-MNIST, CIFAR-10, COIL-100, MVTec AD, and 2 medical datasets (Head CT (hemorrhage) and Brain MRI Images for Brain Tumor Detection).

Datasets MNIST, Fashion-MNIST, and CIFAR-10 will be downloaded by Torchvision. You have to download COIL-100, MVTec AD, Head CT (hemorrhage), and Brain MRI Images for Brain Tumor Detection, and unpack them into the Dataset folder.

Train the Model:

Start the training using the following command. The checkpoints will be saved in the folder outputs/{dataset_name}/{normal_class}/checkpoints.

Train parameters such as dataset_name, normal_class, batch_size and etc. can be specified in configs/config_train.yaml.

python train.py --config configs/config_train.yaml

Test the Trained Model:

Test parameters can be specified in configs/config_test.yaml.

python test.py --config configs/config_test.yaml