Deblending and Classifying Astronomical Sources with Mask R-CNN Deep Learning. Burke, C. J., Aleo, P. D., Chen, Y., Liu, X., Peterson, J. R., Sembroski, G. H., & Yao-Yu Lin, J. arXiv e-prints, 1908:arXiv:1908.02748, August, 2019.
Deblending and Classifying Astronomical Sources with Mask R-CNN Deep Learning [link]Paper  abstract   bibtex   
We apply a new deep learning technique to detect, classify, and deblend sources in multi-band astronomical images. We train and evaluate the performance of an artificial neural network built on the Mask R-CNN image processing framework, a general code for efficient object detection, classification, and instance segmentation. After evaluating the performance of our network against simulated ground truth images for star and galaxy classes, we find a purity of 92% at 80% completeness for stars and a purity of 98% at 80% completeness for galaxies in a typical field with \${\textbackslash}sim30\$ galaxies/arcmin\${\textasciicircum}2\$. We investigate the deblending capability of our code, and find that clean deblends are handled robustly during object masking, even for significantly blended sources. This technique, or extensions using similar network architectures, may be applied to current and future deep imaging surveys such as LSST and WFIRST. Our code, Astro R-CNN, is publicly available at https://github.com/burke86/astro_rcnn.
@article{burke_deblending_2019,
	title = {Deblending and {Classifying} {Astronomical} {Sources} with {Mask} {R}-{CNN} {Deep} {Learning}},
	volume = {1908},
	url = {http://adsabs.harvard.edu/abs/2019arXiv190802748B},
	abstract = {We apply a new deep learning technique to detect, classify, and deblend sources in multi-band astronomical images. We train and evaluate the performance of an artificial neural network built on the Mask R-CNN image processing framework, a general code for efficient object
detection, classification, and instance segmentation. After evaluating the performance of our network against simulated ground truth images for star and galaxy classes, we find a purity of 92\% at 80\% completeness for stars and a purity of 98\% at 80\% completeness for galaxies in a typical field with \${\textbackslash}sim30\$ galaxies/arcmin\${\textasciicircum}2\$. We investigate the deblending capability of our code, and find that clean deblends are handled robustly during object masking, even for significantly blended sources. This technique, or extensions using similar network architectures, may be applied to current and future deep imaging surveys such as LSST and WFIRST. Our code, Astro R-CNN, is publicly available at
https://github.com/burke86/astro\_rcnn.},
	urldate = {2019-08-08},
	journal = {arXiv e-prints},
	author = {Burke, Colin J. and Aleo, Patrick D. and Chen, Yu-Ching and Liu, Xin and Peterson, John R. and Sembroski, Glenn H. and Yao-Yu Lin, Joshua},
	month = aug,
	year = {2019},
	keywords = {Astrophysics - Astrophysics of Galaxies, Astrophysics - Instrumentation and Methods for Astrophysics},
	pages = {arXiv:1908.02748},
}

Downloads: 0