Exoplanet Imaging Data Challenge

Slide 1

Slide 1 text

Exoplanet Imaging Data Challenge Carlos Alberto Gomez Gonzalez Global Challenges Science Week. June3, 2019

Slide 2

Slide 2 text

{insert research topic here} Data Challenge Carlos Alberto Gomez Gonzalez Global Challenges Science Week. June3, 2019

Slide 3

Slide 3 text

Exoplanet Imaging Data Challenge Planet: a celestial body moving in an elliptical orbit round the Sun

Slide 4

Slide 4 text

Exo from extrasolar (outside of the Solar System)

Slide 5

Slide 5 text

Exoplanet Imaging Data Challenge

Slide 6

Slide 6 text

Data challenges • Problem to solve • Benchmark data • Metrics • Platform for hosting the competition • Motivate people to participate (*)

Slide 7

Slide 7 text

Credit: Martin Vargic

Slide 8

Slide 8 text

Credit: NASA, http://planetquest.jpl.nasa.gov High-contrast imaging Lighthouse -> Star Fireﬂy -> exoplanet

Slide 9

Slide 9 text

Credit: NASA, http://planetquest.jpl.nasa.gov High-contrast imaging Lighthouse -> Star Fireﬂy -> exoplanet

Slide 10

Slide 10 text

Credit: NASA, http://planetquest.jpl.nasa.gov High-contrast imaging Lighthouse -> Star Fireﬂy -> exoplanet Stars never shut off !!!

Slide 11

Slide 11 text

Debris disk Beta Pictoris (occulted)

Slide 12

Slide 12 text

Yep, this is the exoplanet (Beta Pictoris b)

Slide 13

Slide 13 text

4k conﬁrmed exoplanets but only ~1% observed!

Slide 14

Slide 14 text

HR8799, Marois et al. 2010 20 AU 0.5” Konopacky et al. 2013 Bowler 2016 Milli et al. 2016

Slide 15

Slide 15 text

Multi-dimensional arrays ( ) exoplanet t, λ, x, y 3D 4D

Slide 16

Slide 16 text

Fake planet (low contrast) Signal and noise

Slide 17

Slide 17 text

spatially rescaled images t0 t0 with an integral ﬁeld spectrograph Signal and noise

Slide 18

Slide 18 text

Image processing pipeline Getting rid of the noise…

Slide 19

Slide 19 text

Algorithms 2006 - 2019 • Median frame subtraction • Pairwise subtraction (GRAPHIC) • Least squares image combination (LOCI) • Low-rank approximations (PCA, NMF) • Low-rank plus sparse decompositions (LLSG) • MLE & Matched ﬁltering (ANDROMEDA, KLIP-FM) • Supervised detection with deep neural networks (SODINN)

Slide 20

Slide 20 text

No content

Slide 21

Slide 21 text

No content

Slide 22

Slide 22 text

Data • Lack of benchmark data in the HCI community • Requested datasets from the most important HCI instruments on Earth • Heterogeneous in terms of quality and characteristics • Synthetic companions were injected

Slide 23

Slide 23 text

No content

Slide 24

Slide 24 text

Submissions

Slide 25

Slide 25 text

Blob counting

Slide 26

Slide 26 text

Metrics

Slide 27

Slide 27 text

No content

Slide 28

Slide 28 text

Exercise of Openness

Slide 29

Slide 29 text

Thank you carlgogo.github.io/ github.com/carlgogo/ speakerdeck.com/carlgogo/ linkedin.com/in/carlgogo/