Institut national de recherche scientifique français Univerité Pierre et Marie Curie Université Paris Diderot - Paris 7

Deep learning-based focal-plane wavefront sensing : principle, simulations, and in-lab tests

lundi 24 avril 2023, par Maxime QUESNEL (Université de Liège)

Jeudi 27 avril 2023 à 16h00 , Lieu : Salle de rĂ©union du 1er Ă©tage du bâtiment 16 et visioconfĂ©rence

High-contrast imaging instruments are today primarily limited by non-common path aberrations appearing between the scientific and wavefront sensing arms. These aberrations can produce quasi-static speckles in science images that are difficult to distinguish from exoplanet signatures. With the help of recent advances in deep learning, we have implemented convolutional neural networks (CNN) to estimate pupil-plane phase aberrations from point spread functions (PSF). In this talk, I will show results with simulations obtained behind a vortex coronagraph, exploiting its properties to provide an alternative type of phase diversity with a 100% science duty cycle. I will also introduce an autoencoder-based method, that uses a deep CNN as the encoder and a differentiable simulator of the instrument as the decoder. This enforces the latent space to represent phase aberrations, and because the approach is unsupervised, it is not necessary to know the true aberrations to train the models. This is particularly promising for on-sky applications, and results on laboratory data using the Subaru/SCExAO instrument are first presented.

We will meet in the meeting room on the first floor of building 16 or on zoom.