Deep Transfer Learning for Blended Source Identification in Galaxy Survey Data

Abstract

We present BlendHunter, a proof-of-concept for a deep transfer learning based approach for the automated and robust identification of blended sources in galaxy survey data. We take the VGG-16 network with pre-trained convolutional layers and train the fully connected layers on parametric models of COSMOS images. We test the efficacy of the transfer learning by taking the weights learned on the parametric models and using them to identify blends in more realistic CFIS-like images. We compare the performance of this method to SEP (a Python implementation of SExtractor) as function of noise level and the separation between sources. We find that BlendHunter outperforms SEP by $\sim 15\%$ in terms of classification accuracy for close blends ($<10$ pixel separation between sources) regardless of the noise level used for training. Additionally, the method provides consistent results to SEP for distant blends ($\ge 10$ pixel separation between sources) provided the network is trained on data with a relatively close noise standard deviation to the target images. The code and data have been made publicly available to ensure the reproducibility of the results.

arXiv comment9 pages, 6 figures, accepted for publication in A&A

References in corpus (22)

• Adam: A Method for Stochastic Optimization
• The Cosmic Evolution Survey (COSMOS) -- Overview
• The Hyper Suprime-Cam Software Pipeline
• Weak lensing for precision cosmology
• The effects of seeing on Sersic profiles. II. The Moffat PSF
• Deep Transfer Learning: A new deep learning glitch classification method for advanced LIGO
• Glitch Classification and Clustering for LIGO with Deep Transfer Learning
• A Taxonomy of Deep Convolutional Neural Nets for Computer Vision
• The Third Gravitational Lensing Accuracy Testing (GREAT3) Challenge Handbook
• Galaxy Morphology Classification with Deep Convolutional Neural Networks
• Radio Galaxy Zoo: ClaRAN - A Deep Learning Classifier for Radio Morphologies
• SCARLET: Source separation in multi-band images by Constrained Matrix Factorization
• Precision simulation of ground-based lensing data using observations from space
• A study of the sensitivity of shape measurements to the input parameters of weak lensing image simulations
• Euclid Preparation IV. Impact of undetected galaxies on weak-lensing shear measurements
• Deep Transfer Learning for Star Cluster Classification: I. Application to the PHANGS-HST Survey
• Dark Energy Survey Year 1 Results: The Impact of Galaxy Neighbours on Weak Lensing Cosmology with im3shape
• Deblending galaxy superpositions with branched generative adversarial networks
• Deblending galaxies with Variational Autoencoders: a joint multi-band, multi-instrument approach
• A bias in cosmic shear from galaxy selection: results from ray-tracing simulations
• Multi-band morpho-Spectral Component Analysis Deblending Tool (MuSCADeT): Deblending colourful objects
• Classification of Planetary Nebulae through Deep Transfer Learning

Cited by in corpus (3)

• ShapePipe: A modular weak-lensing processing and analysis pipeline
• CzSL: Learning from citizen science, experts and unlabelled data in astronomical image classification
• Breaking the degeneracy in stellar spectral classification from single wide-band images