AMLD EPFL 2025: AI & Imaging: Next Frontiers in Image Analysis

The track will focus on the transformative role of machine learning (ML) in image analysis, a field where ML has been instrumental since its early days. ML algorithms power applications such as detection, classification, segmentation, 3D reconstruction, and image generation, and today, state-of-the-art techniques in image analysis rely heavily on these tools. As an interdisciplinary and rapidly evolving area, image analysis benefits from several exciting trends, promising advancements in both academic research and industry practices.

A key focus will be on the scientific applications of image analysis, which span various fields from earth observation to medical imaging. While the image analysis tasks tackled in these fields have strong domain specificities, the core algorithms are often shared across research fields. Hence there is significant potential for innovation across the boundaries of individual disciplines. This track aims to showcase the latest high-impact developments in the image analysis area, explore open research questions, and highlight where the academic community is headed.

On the technological front, ML methods have matured and are now deeply integrated into image analysis pipelines. However, transitioning from proof of concept to widely-usable tools remains a complex journey. The track will feature inspiring work from both academia and industry, spotlighting efforts that not only push the boundaries of image analysis but also reimagine how we approach the challenge of bringing new developments to practical use. This includes exploring questions about the future of image analysis, and reflecting on the opportunities for novel human-machine interactions offered by advanced ML.Key takeaways for participants include gaining insights into the latest advances in ML for image analysis, discovering synergies between disciplines, and fostering connections across the academic and industry communities.

‍Prerequisites

‍We expect participants to have basic knowledge of machine learning and image analysis. While it could be beneficial to know about neural network architectures and current image analysis softwares, we will make sure speakers give a pedagogical introduction for a broad audience. No special equipment is required for this track.

‍

‍Session Chairs

‍Laurène Donati, Switzerland

Jonathan DONG, Switzerland

Mathieu Salzmann, Switzerland

Virginie Uhlmann, Switzerland

‍

‍Line Up

‍14:00 - 14:30 Amir Zamir: Multimodal learning

14:30 - 15:00 Maria Vakalopoulou: AI in Precision Medicine and Medical Imaging

15:00 - 15:30 Paolo Favaro: Towards Building Controllable World Models

15:30 - 16:00 Coffee break

16:00 - 16:30 Ida-Maria SINTORN: AI in biomedical image analysis - bridging the gap between curated proof of concept promises and real world deployment

16:30 - 17:00 Devis Tuia: Mapping and monitoring corals with cheap cameras and machine learning

17:00 - 17:30 Tingying Peng