digiecg_lead_detection_and_signal_segmentation_models

This archive contains two trained models used in the DigiECG pipeline for automated digitization of printed and mobile-captured 12-lead ECG records. The first model localizes the 12 ECG leads by predicting one bounding box per lead across diverse layouts, rotations, and imaging conditions. The second model performs pixel-level separation of ECG signal from background within each detected lead region, enabling accurate waveform extraction for downstream signal reconstruction and analysis.