autosourceID project:
Automatized Identification
of Astrophysical Objects

autosourceID: About

AutoSourceID is a framework for quickly locating and identifying sources in astronomical images.

AutoSourceID currently works for gamma rays and optical images and consists of a chain of neural networks and other tools to rapidly create a catalogue of sources from an image including uncertainties. The algorithm can be split into the following steps: mask generation, source localization, and finally feature extraction and classification.

Fiorenzo Stoppa, Paul Vreeswijk, Steven Bloemen, Saptashwa Bhattacharyya, Sascha Caron, Gu├░laugur J├│hannesson, Roberto Ruiz de Austri, Chris van den Oetelaar, Gabrijela Zaharijas, Paul. J. Groot, Eric Cator, Gijs Nelemans, Boris Panes, Christopher Eckner, Luc Hendriks, Klaas Dijkstra

Please let us know if you like to join / contribute.


AutoSourceID's gamma ray datasets can be found at this Github link
AutoSourceID-Light's datasets can be found at this
Zenodo link

autosourceID: code

AutoSourceID's code can be found at this
Github link
AutoSourceID-Light's code can be found at this
Github link

autosourceID: publications

  • Identification of point sources in gamma rays using U-shaped convolutional neural networks and a data challenge (2021)
  • AutoSourceID-light. Fast Optical Source Localization via U-Net and Laplacian of Gaussian (2022)