BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Pushing the limits of exoplanet discovery via direct imaging with 
 deep learning - Dr Piero Coronica\, UIS\, University of Cambridge
DTSTART:20191009T120000Z
DTEND:20191009T130000Z
UID:TALK131422@talks.cam.ac.uk
CONTACT:Jeffrey Salmond
DESCRIPTION:Exoplanets are planets located outside the solar system. One m
 ethod to detect these distant worlds is through the direct detection of th
 eir thermal emission. The so-called direct imaging technique is suitable f
 or observing young planets far from their star.\n\nDue to the star emissio
 ns\, these are very low signal-to-noise-ratio measurements. Moreover\, the
  limited and highly unbalanced ground truth hinders the use of supervised 
 learning approaches to automatically detect planets signals in the images.
 \n\nIn this talk\, we show how to bypass the scarcity of real data by trai
 ning a Generative Adversarial Network. The synthetic images produced by th
 e generative model can be assumed to not contain any planet and are augmen
 ted by artificially injecting planets signals. The data obtained are not j
 ust labeled but\, for the positive samples\, the exact position of the obj
 ect to detect is known. CNN detectors trained on this synthetic dataset ex
 hibit good predictive performance and\, on real data\, the models can re-c
 onfirm bright sources detection. In this sense\, the above technique shows
  the potential to go beyond the current state of the art in exoplanet disc
 overy via direct imaging.
LOCATION:Rayleigh Seminar Room\, Maxwell Centre\, Cavendish Laboratory
END:VEVENT
END:VCALENDAR