1st Workshop on From Scene Understanding to Human Modeling

Overview

This workshop explores advancements in human modeling and contextual scene analysis within deep learning, crucial for interpreting complex visual environments across various applications like robotics and medical imaging. It emphasizes the accurate modeling of humans and the intricate relationships between all entities in a scene. The workshop will delve into sophisticated human representations (motion, physical plausibility, inferred states) for improved action recognition and behavior prediction, alongside techniques for understanding overall scene context (scene graph prediction, temporal dynamics). A key focus is the interplay between these areas: how detailed human models enhance scene understanding, and how broader scene context refines human modeling. It also highlights emerging datasets and generalized foundation models for jointly reasoning about human behavior and the environment.

 Topics

• Holistic Scene Understanding

• Human Modeling for Scene Understanding

• Action Recognition

• Scene Graph Prediction

• Temporal Scene Understanding

• Foundation Models for Integrated Understanding

• Applications in Specialized Domains

• New Datasets

• etc

Important Dates and Details

• Apply to be part of the Program Committee by: TBA using this link.

• Paper submission deadline: 25 August 2025 11 September 2025

• Submission through CMT: https://cmt3.research.microsoft.com/SUtoHM2025

• Workshop: 27 November 2025