Morpheus: A Deep Learning Framework For Pixel-Level Analysis of Astronomical Image Data
Fine-grain morphological classifications at astronomical scale.
I work on the development and application of novel computer vision techniques for the precision and scale of astronomical image data.
Astronomical data is richly complex and presents unqiue challenges for computer vision techniques. Aspects of astronomical sources such as high dynamic ranges, varying degrees opacity, and a variety of scales make applying off the shelf techniques difficult. Coupled with the volume of data that upcoming telescopes are expected to produce, new techniques need to be developed to enable astronomers to effectively perform their science. I am co-supervised by Professor Brant Robertson in the Computational Astrophysics Research Group and Professor Roberto Manduchi in the UCSC Computer Vision Lab.
Active Work
FitsMap
A tool for displaying astronomical images and their associated catalogs.
Morpheus Core
The core framework code to scale up per-pixel machine learning methods to large astronomical images.
Partial-Attribution Instance Segmentation for Astronomical Source Detection and Deblending
A new technique for the detection and deblending of astronomical sources using deep learning.