Ducktective: CLI anomaly detector

This project is a small yet powerful command-line interface (CLI) application designed for AI-based anomaly detection, specifically tailored for industrial use. It leverages autoencoders to identify outliers by training exclusively on images of "good" samples.

The application trains a model on the provided images of normal (good) samples, learning their key features and patterns. When supplied with new sample images, the application reconstructs each image using the trained deep learning model and calculates the difference between the original and reconstructed images. Based on this difference, it classifies the sample as either "good" or an "outlier.

Key features:

  • Industrial use case

    Good for quality control in manufacturing and similar industries.

  • Reporting

    Automatically generates reports that include tables and visualizations of the detected anomalies.

  • Metrics & statistics

    Offers training metrics and detection statistics for performance evaluation.

  • Streamlined operation

    User-friendly CLI ensures ease of integration into existing workflows.