Gedas Bertasius

Gedas Bertasius is an Assistant Professor in the Department of Computer Science at the University of North Carolina at Chapel Hill. Previously, he served as a postdoctoral researcher at Meta AI (Facebook AI) and earned his PhD in Computer Science from the University of Pennsylvania. His academic journey began with a bachelor’s degree in Computer Science from Dartmouth College.

Dr. Bertasius specializes in computer vision and machine learning with specific interests in:

• Video understanding
• First-person vision (egocentric vision)
• Human behavior modeling
• Multimodal deep learning
• Transfer learning
• Computer vision for sports analytics
• Video+robotics integration

His research produces practical frameworks like Video ReCap for hierarchical captioning of long videos, SiLVR for language-based video reasoning, and BASKET for fine-grained skill estimation. He focuses on developing models that can process videos across multiple temporal granularities while maintaining computational efficiency.

Key research themes in his work include:

• Recursive video processing architectures
• Space-time attention mechanisms
• Generative video modeling
• LLM integration with vision systems
• 3D-aware representation learning
• Continual learning for video QA

He has received notable recognition, including:

• CVPR 2024 Egocentric Vision (EgoVis) Distinguished Paper Award
• CVPR 2020 Best Paper Award Nomination
• First Place at CVPR 2025 Multi-Discipline Lecture Understanding Workshop

Dr. Bertasius collaborates with prominent researchers like Mohit Bansal and Lorenzo Torresani. His recent publications demonstrate expertise in advancing video-language models, with applications in semantic alignment, temporal grounding, and cross-modal reasoning. For detailed information about his research, publications, and ongoing projects, please visit his official website.