We are proud to present MaxiMask and MaxiTrack, a set of softwares using convolutional neural networks to detect problematic pixels and problematic astronomical images.

MaxiMask has been trained to identify:

   - cosmic rays

   - dead/hot pixels, including dead/hot columns or rows

   - persistence effects

   - satellite trails

   - fringes

   - nebulosity / extended emission

   - saturated stars and associated blooming artifacts

   - diffraction spikes

It was designed from the beginning to be very versatile and can deal equally well with CCD or infrared detectors.It is now extensively used with a broad range of astronomical cameras, including for example CFHT/Megacam, CTIO/Decam, Subaru/HSC, VST/Omegacam.

Fig: Examples of MaxiMask results. Left: real astronomical images augmented with different kinds of artifacts. Right: MaxiMask detection results. Note that stars are detected as well (purple).


The code is available on-line on github.

Try it and send us your comments!



Maximask performs very well on Starlink satellite trails plaguing astronomical images. The DECam image below shows an example that received a lot of attention in the news worldwide. On the left the original image, and on the right the satellite trails detected by MaxiMask.

The figure on the right shows an example of an image with guiding or tracking problem identified by MaxiTrack. Such images happen when the telescope control fails, when earthquake shake the observatory (many professional observatories are in seismic areas), or when the observer is pointing at a solar system moving object.


The code is also available on-line on github.




Find out more in our publication:

Paillassa, Bertin & Bouy, 2020, A&A, 634, 48



© Last Update: 03-11-2020 by H. Bouy