AstroML Service

The AstroML service has been developed to improve source identification, classification and characterization of sources in large-scale radio surveys. This release includes a deep convolutional neural network based on an instance segmentation framework (R-CNN) and the Tiramisu network that allow both detection and classification of radio compact sources, radio galaxies with extended morphology and sidelobe imaging artefacts. It can work either as a standalone source finder, or as a classifier stage applied to source finders catalogue outputs of the CAESAR Service.

Provider

The AstroML Service is provided by INAF-Istituto Nazionale di Astrofisica

Getting Started

STEP 1) AI Gateway Login

  • Navigate to http://ai-gateway.neanias.eu
  • Click the ‘SignIn’ button
  • Click on ‘Neanias Identity Provider’
  • Authenticate using valid credentials (Microsoft or Google)

STEP 2) Spawn AstroML Environment

  • Select ASTRO ML environment and click the ‘Spawn’ button
  • Navigate inside the directory named “example_astroml”
  • Open the ‘RunDetection.ipynb’ sample notebook
  • Run the notebook cells by selecting ‘Cell’ -> ‘Run All’
  • Test with your image: Upload your “.fits” image in data “folder”. Change the name of image into the notebook in line 2. Run the notebook as indicated in the previous step.

STEP 3) Spawn Tiramisu Environment

  • Select TIRAMISU environment and click the ‘Spawn’ button
  • Navigate inside the directory named “example_tiramisu”
  • Open the ‘test.ipynb’ sample notebook
  • Run the notebook cells by selecting ‘Cell’ -> ‘Run All’

Please note: To switch between different environments you will have to click on the “Control Panel” button, then stop your current runng server and start a new environment.

Availability

NEANIAS AstroML Service: https://caesar.dev.neanias.eu/caesar/api/v1.0/app/sfinder-nn/describe

NEANIAS AstroML Jupyther Notebooks available from the C3.1 AI Gateway: http://ai-gateway.neanias.eu