Fermi-LAT Source Classification

Project developed by Pascal Napoli for the Computing Methods For Experimental Physics And Data Analysis exam.

This project implements a complete data analysis pipeline for the automatic classification of astrophysical sources from the Fermi-LAT 4FGL catalog. The primary goal is to distinguish between the two largest classes of gamma-ray emitters—Active Galactic Nuclei (AGN) and Pulsars—and to predict classifications for unassociated sources based on their spatial, temporal, and spectral characteristics.

The core of the project is a Deep Neural Network (DNN) trained to analyze these features with high accuracy.

An automatically generated interactive map displaying the distribution of classified sources is available.

• Visualization: Uses Plotly to project sources onto a Mollweide projection using Galactic Coordinates.
• Automation: A GitHub Action automatically regenerates and deploys the map to GitHub Pages whenever the dataset or map script is updated.

👉 Click here to view the Interactive Map with All Sources

👉 Click here to view the Interactive Map with Only Not Associated Sources

Complete documentation for the project is available here: 👉 Read the Documentation

• dnn/: Contains Jupyter Notebooks, training scripts, and the model architecture.
• map/: Scripts for generating the interactive HTML map.
• fits_import/: Modules for processing raw FITS files into CSV format.
• imports/: Custom utility modules used across the project.
• files/: Contains the raw dataset files.
• .github/workflows/: CI/CD workflows for automation.

All necessary dependencies (e.g., astropy, tensorflow, plotly, pandas) are listed in the requirements.txt file.

To install the environment:

pip install -r requirements.txt