Boundary-Aware Multi-task Network for Aortic Ring Segmentation and Landmark Detection

This dataset contains fluoroscopic TAVI frames annotated for two tasks: binary segmentation of the aortic root and localization of four anatomical landmarks (`AA1`, `AA2`, `STJ1`, `STJ2`). The annotations are provided in Supervisely JSON format and are organized on a per-patient basis. The dataset can be used for multitask training of BAMNet as well as for derived YOLO object detection, segmentation, and keypoint benchmarks.