Python - Anomaly Detection in Agri-Food Supply Chain

In this study, the objective was to conduct a comparative analysis for anomaly detection using two different unsupervised techniques: an Autoencoder and a Self-Organizing Map (SOM).
The study highlights the complementary strenghts of the models:
The Autoencoder captures the structural deviations and the SOM focuses on color quantization and this makes the integration of both of them a natural recommendation for an holistc system.

Check it out at my github repo.

python computer-vision anomaly-detection pytorch tensorflow autoencoder self-organizing-map machine-learning deep-learining

[input, reconstruction, difference (input-reconstruction), heatmap]