mo-anomaly-detection

An multi objective anomaly detection project using deep learning and PyTorch, designed to identify structural and counting anomalies in images.

Table of Contents

• Installation
• Usage
  • Training
  • Testing
  • Inference
• Project Structure
• Features
• Results
• Contributing
• License

Installation

Requirements

• Python >= 3.6
• PyTorch == 2.4.1+cu121
• Torchvision == 0.19.1+cu121
• CUDA 12.1 (for GPU support)

Setup Instructions

1. Clone the repository:

   git clone https://github.com/hafizarslanamjad/mo-anomaly-detection.git
   cd mo-anomaly-detection

2. Create and activate a virtual environment:

   python -m venv venv
   source venv/bin/activate

3. Install dependencies:

   please see setup.py file

4. Install PyTorch with CUDA support (if required):

   pip install torch==2.4.1+cu121 torchvision==0.19.1+cu121 --extra-index-url https://download.pytorch.org/whl/cu121

Usage

Training

To train the anomaly detection model, run:

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

Testing

To evaluate the model on the test set:

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

File will be provided soon

Inference

To perform anomaly detection on new images:

python inference.py --input /path/to/image

File will be provided soon

Note: Adjust paths as needed, and make sure the images follow the directory structure specified in the configuration files.

Project Structure

mo-anomaly-detection/
├── checkpoints/       # Model checkpoints
├── dataset/           # Training/testing data
├── models/            # Model architectures
├── utils/             # Helper functions
├── venv/              # Virtual environment (not included in Git)
├── train.py           # Training script
├── test.py            # Testing script
├── inference.py       # Inference script
├── configs/           # Configuration files
└── README.md          # Project documentation

Checkpoints and Dataset will be made available soon!

Features

• Convolutional Autoencoder with Attention Mechanisms
• Custom dataset processing for anomaly detection
• Configurable training, testing, and inference pipelines

Results

The model achieves __ __ in unsupervised multi objective anomaly detection. Below are sample results:

• MEASURE 1: TO BE POSTED
• MEASURE 2: TO BE POSTED

Contributing

Contributions are welcome! Please open an issue or submit a pull request with improvements or bug fixes.

License

This project is licensed under the MIT License - see the LICENSE file for details.