Tails: Chasing Comets with the Zwicky Transient Facility and Deep Learning

Abstract

We present Tails, an open-source deep-learning framework for the identification and localization of comets in the image data of the Zwicky Transient Facility (ZTF), a robotic optical time-domain survey currently in operation at the Palomar Observatory in California, USA. Tails employs a custom EfficientDet-based architecture and is capable of finding comets in single images in near real time, rather than requiring multiple epochs as with traditional methods. The system achieves state-of-the-art performance with 99% recall, 0.01% false positive rate, and 1-2 pixel root mean square error in the predicted position. We report the initial results of the Tails efficiency evaluation in a production setting on the data of the ZTF Twilight survey, including the first AI-assisted discovery of a comet (C/2020 T2) and the recovery of a comet (P/2016 J3 = P/2021 A3).

References in corpus (23)

• Array Programming with NumPy
• TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems
• Deep Residual Learning for Image Recognition
• Binary companions of evolved stars in APOGEE DR14: Search method and catalog of ~5,000 companions
• The Zwicky Transient Facility: System Overview, Performance, and First Results
• The Zwicky Transient Facility: Data Processing, Products, and Archive
• ATLAS: A High-Cadence All-Sky Survey System
• The Zwicky Transient Facility: Science Objectives
• EfficientDet: Scalable and Efficient Object Detection
• The Zwicky Transient Facility: Surveys and Scheduler
• The Zwicky Transient Facility: Observing System
• Proper image subtraction - optimal transient detection, photometry and hypothesis testing
• Real-bogus classification for the Zwicky Transient Facility using deep learning
• The GROWTH Marshal: A Dynamic Science Portal for Time-Domain Astronomy
• Initial characterization of interstellar comet 2I/Borisov
• An Observational Upper Limit on the Interstellar Number Density of Asteroids and Comets
• Interstellar Interlopers: Number Density and Origins of 'Oumuamua-like Objects
• Detection of CN gas in Interstellar Object 2I/Borisov
• Characterization of the Nucleus, Morphology and Activity of Interstellar Comet 2I/Borisov by Optical and Near-Infrared GROWTH, Apache Point, IRTF, ZTF and Keck Observations
• Planetary Radio Interferometry and Doppler Experiment (PRIDE) technique: A test case of the Mars Express Phobos fly-by
• A Twilight Search for Atiras, Vatiras and Co-orbital Asteroids: Preliminary Results
• The Performance of the Robo-AO Laser Guide Star Adaptive Optics System at the Kitt Peak 2.1-m Telescope
• Preliminary estimates of the Zwicky Transient Facility 'Ayló'chaxnim asteroid population completeness

Cited by in corpus (5)

• A data science platform to enable time-domain astronomy
• Discovering Faint and High Apparent Motion Rate Near-Earth Asteroids Using A Deep Learning Program
• Towards Asteroid Detection in Microlensing Surveys with Deep Learning
• Optimization of Artificial Neural Networks models applied to the identification of images of asteroids' resonant arguments
• Using neural networks to model Main Belt Asteroid albedos as a function of their proper orbital elements