Derived Data and Analysis Records Supporting "Deep Learning-Guided Energy-Augmented Anisotropic Frangi Filtering for Antarctic Ice-Shelf Damage Detection"

This dataset contains the derived numerical results and analysis records supporting the manuscript “Deep Learning-Guided Energy-Augmented Anisotropic Frangi Filtering for Antarctic Ice-Shelf Damage Detection”. The deposit includes multiseed U-Net and SegFormer evaluation results, fixed-pipeline YOLO-EAAF results, aggregated 30-m and common 300-m performance metrics, 300-m occupancy-threshold sensitivity results, PIG component-ablation results, PIG-only hysteresis-threshold sensitivity data, and public-safe records of the model-selection, frozen-inference, and evaluation-domain protocols. Original Landsat, Sentinel-1, and Sentinel-2 imagery is not duplicated in this deposit because the source imagery is publicly available through Google Earth Engine and the corresponding satellite data archives.