NA-SODINN: a deep learning algorithm for exoplanet image detection based on residual noise regimes

Abstract

Supervised deep learning was recently introduced in high-contrast imaging (HCI) through the SODINN algorithm, a convolutional neural network designed for exoplanet detection in angular differential imaging (ADI) datasets. The benchmarking of HCI algorithms within the Exoplanet Imaging Data Challenge (EIDC) showed that (i) SODINN can produce a high number of false positives in the final detection maps, and (ii) algorithms processing images in a more local manner perform better. This work aims to improve the SODINN detection performance by introducing new local processing approaches and adapting its learning process accordingly. We propose NA-SODINN, a new deep learning binary classifier based on a convolutional neural network (CNN) that better captures image noise correlations in ADI-processed frames by identifying noise regimes. Our new approach was tested against its predecessor, as well as two SODINN-based hybrid models and a more standard annular-PCA approach, through local receiving operating characteristics (ROC) analysis of ADI sequences from the VLT/SPHERE and Keck/NIRC-2 instruments. Results show that NA-SODINN enhances SODINN in both sensitivity and specificity, especially in the speckle-dominated noise regime. NA-SODINN is also benchmarked against the complete set of submitted detection algorithms in EIDC, in which we show that its final detection score matches or outperforms the most powerful detection algorithms.Throughout the supervised machine learning case, this study illustrates and reinforces the importance of adapting the task of detection to the local content of processed images.

arXiv commentA&A in press

References in corpus (34)

• Direct Imaging of Multiple Planets Orbiting the Star HR 8799
• Angular Differential Imaging: a Powerful High-Contrast Imaging Technique
• Images of a fourth planet orbiting HR 8799
• Detection and Characterization of Exoplanets and Disks using Projections on Karhunen-Loeve Eigenimages
• Discovery of a planetary-mass companion within the gap of the transition disk around PDS 70
• SPHERE: the exoplanet imager for the Very Large Telescope
• A new algorithm for point spread function subtraction in high-contrast imaging: a demonstration with angular differential imaging
• Fundamental limitations of high contrast imaging set by small sample statistics
• PynPoint: An Image Processing Package for Finding Exoplanets
• Apodized Pupil Lyot Coronagraphs for Arbitrary Telescope Apertures
• Discovery of a probable 4-5 Jupiter-mass exoplanet to HD 95086 by direct-imaging
• VIP: Vortex Image Processing package for high-contrast direct imaging
• Discovery of a warm, dusty giant planet around HIP65426
• The Vector Vortex Coronagraph: Laboratory Results and First Light at Palomar Observatory
• Direct exoplanet detection and characterization using the ANDROMEDA method: Performance on VLT/NaCo data
• Speckle noise and dynamic range in coronagraphic images
• Temporal Evolution of Coronagraphic Dynamic Range, and Constraints on Companions to Vega
• Non-negative Matrix Factorization: Robust Extraction of Extended Structures
• Improving and Assessing Planet Sensitivity of the GPI Exoplanet Survey with a Forward Model Matched Filter
• Searching for companions down to 2 AU from beta Pictoris using the L'-band AGPM coronagraph on VLT/NACO
• Confidence Level and Sensitivity Limits in High Contrast Imaging
• Direct Imaging Discovery of a Jovian Exoplanet Within a Triple Star System
• Low-rank plus sparse decomposition for exoplanet detection in direct-imaging ADI sequences. The LLSG algorithm
• The W. M. Keck Observatory infrared vortex coronagraph and a first image of HIP79124 B
• GPI PSF subtraction with TLOCI: the next evolution in exoplanet/disk high-contrast imaging
• Discovery of a directly imaged planet to the young solar analog YSES 2
• Improving Signal to Noise in the Direct Imaging of Exoplanets and Circumstellar Disks
• Supervised detection of exoplanets in high-contrast imaging sequences
• A New Standard for Assessing the Performance of High Contrast Imaging Systems
• TRAP: A temporal systematics model for improved direct detection of exoplanets at small angular separations
• STIM map: detection map for exoplanets imaging beyond asymptotic Gaussian residual speckle noise
• RSM detection map for direct exoplanet detection in ADI sequences
• Pushing the Limits of Exoplanet Discovery via Direct Imaging with Deep Learning
• Auto-RSM: an automated parameter-selection algorithm for the RSM map exoplanet detection algorithm

Cited by in corpus (3)

• Use the 4S (Signal-Safe Speckle Subtraction): Explainable Machine Learning reveals the Giant Exoplanet AF Lep b in High-Contrast Imaging Data from 2011
• Machine learning for exoplanet detection in high-contrast spectroscopy Combining cross correlation maps and deep learning on medium-resolution integral-field spectra
• An Alternating Minimization Algorithm with Trajectory for Direct Exoplanet Detection -- The AMAT Algorithm