Photometric Analysis of Eclipsing Contact Binaries Using a Multi- branch Convolutional Neural Network for Regression and Dual Classification

This record archives the code and research artifacts accompanying the manuscript “Photometric Analysis of Eclipsing Contact Binaries Using a Multibranch Convolutional Neural Network for Regression and Dual Classification.”
It provides an end-to-end pipeline to (i) generate synthetic over-contact EB light curves with PHOEBE, (ii) train a multitask CNN that performs simultaneous regression (physical parameters), validity detection (optional parameters such as third light and spot terms), and starspot classification, and (iii) evaluate on real TESS light curves.