Synthetic Cometary Dataset for Centroiding Algorithm Selection (Kometra Benchmark Suite)

This dataset provides a standardized, high-fidelity synthetic benchmark suite specifically designed and originally established to evaluate, compare, and select the optimal astrometric centroiding algorithm subsequently adopted within the Kometra software ecosystem. Beyond its primary role in Kometra's development, the suite serves as a general-purpose benchmark for cometary astrometry. The archive is organized into two primary benchmarking subsets: Main Benchmark (300 FITS images): Used for rigorous algorithm evaluation, window-size sensitivity testing, click-offset drift invariance, and final selection of the core centroiding method. Comparative Subset (30 FITS images): A curated evaluation set designed for cross-software testing and performance comparison with third-party astronomical packages. Both subsets include images equally divided into three morphological classes (Baseline, Low_SNR, Asymmetric) generated using physically grounded numerical models, complete with ground-truth coordinates (true_x, true_y) in CSV format.