Pattern Recognition Symposium 2025 in Hochkrimml, Austria
At the end of July, the Pattern Recognition Lab gathered for our annual Pattern Recognition Symposium (PRS) in Hochkrimml, Austria. Five days in the mountains meant plenty of inspiring talks, lively poster sessions, and late-night discussions that continued long after dinner.
The program once again highlighted the breadth of our lab’s research. In medical imaging, we saw advances in Gaussian-based shift-variant CT reconstruction (Chengze Ye), retrieval-based report generation for 3D CT imaging (Lukas Buess), and low-dose CT denoising with self-supervised learning (Yipeng Sun). Other contributions tackled breast MRI screening with AI (Tri Nguyen), stroke onset estimation from CT (Linda Vorberg), and photon-counting CT for atherosclerotic plaque classification (Florian Goldmann).
Speech and language research was equally well represented: improving automatic speech recognition for pathological speech with LLM-based error correction (Abner Hernandez) and multimodal analysis of clinical interviews (Hiu Ching Hung and Tobias Pertlwieser) opened exciting perspectives on bridging technology and communication.
Our utility infrastructure group presented novel approaches to grid protection and operation, including reinforcement learning for coordinated fault protection (Julian Oelhaf) and graph neural networks for congestion management (Changhun Kim). Additional work in this area featured thermal numerical analysis for water networks and forecasting challenges in energy systems.
Computer vision and cross-disciplinary work also had a strong presence: action recognition in artworks (Mathias Zinnen), spatiotemporal learning for calving front segmentation (Marcel Dreier), pre-training hybrid transformer-CNNs for glacier segmentation (Nora Gourmelon), and extended reality preparation for MRI (Aniol Serra Juhé). We also heard about SidewalkGPT, a vision-language system to empower the blind (Hakan Calim), as well as explorations into variational quantum error correction (Nico Meyer).
Poster sessions added further depth, featuring work such as intelligent lesion selection for breast cancer metastases (Melika Qahqaie), multimodal image registration for prostate cancer diagnosis (Lisa Hofbauer), foundation models adapted to X-ray imaging (Joshua Scheuplein), and vision transformers for dense predictions (Siyuan Mei).
What makes PRS unique is how all of this comes together in one setting. Talks, posters, and informal conversations interwove across topics, creating an atmosphere where ideas grew and collaborations took shape. The group photo with the Alps in the background captures the spirit perfectly: science and community side by side 🔬🤝🏔️
A big thank you to everyone who contributed, both onsite and online. We are already looking forward to PRS 2026!



