Connecting 3D and Language
COM2 Level 4
Executive Classroom, COM2-04-02
Abstract:
People communicate about objects, scenes, and spatial relations in the real world using natural language. To endow computational systems with the ability to connect natural language and 3D representations, we need to bridge the gap between the high-level, underspecified constructs of language and the rich geometric details of 3D. In this talk, I will give an overview of recent work on language and 3D and recent trends in 3D content generation from text. I will describe recent projects in my group on aligning text to 3D content, 3D scene generation, and evaluation of natural language grounding in 3D scenes.
Bio:
Angel Chang is an Associate Professor at Simon Fraser University and a Canada CIFAR AI Chair with Amii. She received her Ph.D. in Computer Science from Stanford, where she was part of the Natural Language Processing Group and advised by Chris Manning. Her research connects language to visual and 3D representations, and grounds language for embodied agents in indoor environments. She has worked on synthesizing 3D scenes and shapes from natural language, as well as localizing objects in 3D. Her work has been recognized by awards such as the SGP dataset award (for ShapeNet and ScanNet), and most recently with the 3DV 2025 best paper award (for Omages).