Mahsa Ehsanpour

Papers

Total Citations

H-Index

About

Mahsa Ehsanpour is a leading researcher in computer vision and human behavior understanding, with a focus on spatio-temporal action detection, social group analysis, and human motion forecasting. Her work addresses the critical challenge of interpreting complex, real-world scenes involving multiple people, bridging the gap between controlled laboratory settings and unconstrained environments. Ehsanpour’s major contributions include the creation of the JRDB-Act dataset (49 citations), a large-scale benchmark that enables the simultaneous detection of actions, social groups, and activities in crowded, naturalistic settings—a foundational resource for advancing autonomous systems and surveillance. She also developed TRiPOD, a pioneering framework for jointly forecasting human trajectories and pose dynamics, capturing subtle interaction cues essential for robotics and autonomous driving. Her research has been recognized for its impact on enabling machines to understand and predict human social behavior in the wild, with applications ranging from safe human-robot collaboration to intelligent monitoring systems. Ehsanpour’s work stands out for its integration of action recognition with social context, pushing the boundaries of how AI interprets dynamic, multi-agent environments.