Best Paper Award (Third Place) at the IEEE Intelligent Vehicles Symposium for Panoramic Panoptic Segmentation: Towards Complete Surrounding Understanding via Unsupervised Contrastive Learning (A. Jaus, K. Yang, R. Stiefelhagen) — introducing panoramic panoptic segmentation for holistic 360° scene understanding via the Panoramic Robust Feature (PRF) framework.
Our “zero sugar” tracer-aware data-filtering approach placed 2nd in the Data-Centric Track of the AutoPET III Challenge — a leading international competition (490+ teams) on robust lesion segmentation in whole-body PET/CT across tracers and clinical centers.
1st place in the Scribble-based Track (Filters, Thresholds, and Geodesic Distances) and 6th in the main Laptop Challenge (Revisiting classical segmentation methods) at CVPR 2024. We showed that classical, low-resource methods — thresholding, k-means, interpolation — can match or outperform foundation models like MedSAM across 11 imaging modalities, at a fraction of the compute.
Recipient of the Helmholtz Information & Data Science School for Health scholarship — a structured doctoral program jointly run by KIT, the German Cancer Research Center (DKFZ), and Heidelberg University, bringing together 40+ interdisciplinary research groups at the intersection of data science and health.
