Applied Vision

Organizers

Harald Ganster^*, Florian Kleber^#, Roland Perko^*, Gernot Stübl⁺, Martina Uray^*

^* Joanneum Research
^# Vienna University of Technology
⁺ Profactor GmbH

Important dates

July 1, 2025 Camera Ready Paper deadline

Workshop Description

The Austrian Association for Pattern Recognition (OAGM) invites authors and guests to participate in the Applied Vision Workshop as part of the AIRoV event, held in person at the Alpen-Adria-University Klagenfurt. The workshop provides a platform for presentation and discussion of research progress as well as current projects within the OAGM community. We additionally focus on interdisciplinary research of computer vision and pattern analysis in the contexts of Applied Vision. The aim is to bring together Austrian and nearby located groups, researchers, and students working on interdisciplinary topics for discussion and establishing potential collaborations.

General topics of interest include but are not limited to:

3D vision, stereo, and structure from X
Animal and plant phenotyping
Cultural heritage
Cyber-physical systems
Document analysis
Embedded and mobile computer vision / machine learning
Explainability, ethics, and fairness in computer vision / machine learning
Image denoising, restoration & enhancement
Image, video, and multimodal retrieval
Industrial assistance systems
Inspection and quality assurance
Motion and tracking
Multi- and hyperspectral image analysis
Object/scene detection, recognition, and categorization
Pattern analysis (visual and non-visual data)
Precision agriculture and precision forestry
Remote sensing and earth observation
Sensors and information fusion
Video analysis and event recognition
Vision datasets, benchmarking, and performance evaluation

Program

Invited Talk: On Pretraining for Earth Observation

The increasing volume and variety of Earth observation (EO) data present both opportunities and challenges for developing effective machine learning models. Traditional supervised approaches often fall short due to limited amounts of labeled data and high annotation costs, particularly across diverse sensor modalities. Pretraining has emerged as a powerful strategy to overcome these limitations, yet current trends toward large foundation models raise concerns regarding computational efficiency and task specificity. In our research, we explore task-specific pretraining methods tailored for core EO tasks such as semantic segmentation and change detection. We highlight the potential of targeted pretraining to improve performance and efficiency while advancing the goals of AI democratization, sustainability, and technological sovereignty in EO applications.

Luka Čehovin Zajc is an Assistant Professor at the Faculty of Computer and Information Science, University of Ljubljana. His early research focused on visual object tracking, where he contributed to the development of evaluation protocols and benchmarking practices through the Visual Object Tracking (VOT) Initiative. In recent years, his work has shifted toward Earth observation, with an emphasis on robust deep learning methods and their application to real-world problems in collaboration with experts in remote sensing and geoinformatics.

Workshop Track Schedule for Tuesday July 8, 2025

Time	Paper Title	Authors
13:30–14:10	On Pretraining for Earth Observation (Short – Keynote)	Luka Cehovin*; Blaž Rolih (UL FRI); Filip Wolf (UL FRI)
14:10–14:30	Photogrammetry meets Computer Vision: Simplified LiDAR to Camera Calibration (Full)	Richard Ladstädter; Gerald Lodron; Manfred Klopschitz; Roland Perko (Joanneum Research)*
14:30–14:45	Human Perception for Print Defect Detection (Short)	Lukas Brunner (Profactor)*; Thomas Pönitz (Profactor)
14:45–15:00	Visual emotion recognition for an empathic vehicle (Short)	Raquel Panadero Palenzuela (TU Wien); Michael Hödlmoser (emotion3D GmbH); Niklas Fraissl (TU Wien, FH Technikum)*; Timothy Berens (FH Technikum Wien); Dominik Schörkhuber (TU Wien); Benedikt Salzbrunn (FH Technikum Wien); Sandra Kitting (FH Technikum Wien); Rosen Dimov (emotion3D GmbH); Margrit Gelautz (TU Wien)
15:00–15:30	Coffee Break
15:30–15:45	Vision-based Measurement of Rail Sleeper Vibrations (Short)	Gerald Zauner (FH OÖ, Wels)*
15:45–16:00	Camera Movement Classification in Historical Footage: A Comparative Study of Deep Video Models (Short)	Tingyu Lin (TU Wien)*; Armin Dadras; Florian Kleber; Robert Sablatnig (TU Wien)
16:00–16:20	Assessing the Impact of Binarization for Writer Identification in Greek Papyrus (Full)	Dominic Akt (TU Wien); Marco Peer (HES-SO)*; Florian Kleber (TU Wien)
16:20–16:40	Soft-label based regression for contamination grading with HSI in plastics recycling (Full)	Sergei Vedyshev (Joanneum Research)*; Harald Ganster (Joanneum Research); Malte Jaschik (Joanneum Research);
16:40–16:55	Efficient Automatic Detection of Scanned Body Regions in CT Scans (Short)	Bertram Sabrowsky-Hirsch*; Ahmed Alshenoudy; Stefan Thumfart (RISC Software GmbH); Thomas Potrusil; Johanna Gottwald; Mario Moser (CADS GmbH)

Program Committee

Doris Antensteiner - Austrian Institute of Technology
Christof Bertram - University of Veterinary Medicine
Csaba Beleznai - Austrian Institute of Technology, Austria
Harald Ganster - Joanneum Research
Katharina Hofer-Schmitz - Joanneum Research
Florian Kleber - Vienna University of Technology
Martin Kampel - Vienna University of Technology, Computer Vision Lab
Olaf Kaehler - Joanneum Research
Christoph Lampert - IST Austria
Sead Mustafic - Joanneum Research
Christoph Praschl - University of Applied Sciences Upper Austria
Roland Perko - Joanneum Research
Robert Sablatnig - Vienna University of Technology
Josef Scharinger - Johannes Kepler University Linz
Gernot Stübl - Profactor GmbH
Stefan Thumfart - RISC Software GmbH
Martina Uray - Joanneum Research
Martin Welk - UMIT TIROL
Peter M. Roth - University of Veterinary Medicine, Vienna
Matthias Zeppelzauer - St. Pölten University of Applied Sciences
Gerald Zwettler - University of Applied Sciences Upper Austria

Contact:

Email: appliedVision-airov24@cvl.tuwien.ac.at