VIS goes ACM MM 2025

For the third consecutive year, our institute (VIS) was represented at ACM Multimedia. This time, we contributed to two notable areas at the event in Dublin: the co-organization of a workshop on UAV-based multimedia analysis and achieving first place in the GroMo Challenge on multi-view plant phenotyping.

This year, Fabian Deuser co-organized the 3rd Workshop on UAVs in Multimedia: Capturing the World from a New Perspective (UAVM), helping to shape one of the conference’s thematic tracks. The workshop featured two keynote speakers, Daniel-André Dücker from the Technical University of Munich and Zhedong Zheng from the University of Macau, whose talks addressed underwater robotics, aerial spatial intelligence, and the challenges of autonomous perception in GPS-denied or visually degraded environments. The workshop attracted strong participation, with more than 100 registered teams for the research challenge and 14 accepted papers. UAVM serves as a dedicated forum for drone-based multimedia research, covering aerial image and video processing, machine learning for UAV data, swarm coordination, and UAV-enabled applications. Within ACM Multimedia, it brings together researchers and practitioners working on the challenges and opportunities that arise when UAVs become a central platform for capturing and interpreting large-scale visual data.

Additionally, our institute achieved outstanding research success at the 2025 ACM MM through the GroMo Challenge in precision agriculture. Robin-Nico Kampa presented the study "ViewSparsifier: Killing Redundancy in Multi-View Plant Phenotyping", which introduces a robust approach for handling view redundancy in multi-view plant images. The method combines strategic view selection with Transformer-based feature fusion to extract compact and informative representations from highly overlapping camera perspectives. This approach won first place in both challenge tasks, Plant Age Prediction and Leaf Count Estimation, demonstrating strong generalization across multiple crops and making a significant contribution to scalable, automated plant phenotyping.
The GroMo Workshop focused on advancing smart farming technologies for the future, emphasizing how AI-driven methods can enhance efficiency and sustainability in agricultural research. Discussions explored emerging directions in plant phenotyping, including the integration of generative AI to predict and simulate future plant growth stages based on visual observations. This forward-looking perspective highlights the growing role of machine learning and generative modeling in building autonomous, data-efficient systems for digital agriculture.

Together, VIS’s workshop organization and its winning GroMo contribution illustrate the institute’s active role within the ACM Multimedia community.

The 3rd Workshop on UAVs in Multimedia: Capturing the World from a New Perspective

Tingyu Wang, Yujiao Shi, Fabian Deuser, Shaofei Huang, Guosheng Hu, Si Liu, Zhedong Zheng, Roger Zimmermann

[LINK]

ViewSparsifier: Killing Redundancy in Multi-View Plant Phenotyping

Robin-Nico Kampa, Fabian Deuser, Konrad Habel, Norbert Oswald

[LINK]