Video by Meta Open Source via YouTube
Watch the latest updates from the research community at the Third Joint Egocentric Vision (EgoVis) Workshop at CVPR 2026 in Denver, Colorado! 🚀
This year’s workshop dives deep into how wearable cameras and AR/VR devices leverage AI to understand human behavior, predict object interactions, and map environments in 3D.
0:01 Tushar Nagarajan – Introduction
6:36 Marc Pollefeys – Keynote
36:30 Luigi Seminara – ViterbiPlanNet with
41:08 Tim Engelbracht – Hoi! – A Multimodal Dataset for Force-Grounded, Cross-View Articulated Manipulation
46:12 Olga Loginova – How to Correctly Make Mistakes
51:25 Enrico Pellota – EgoControl: Controllable Egocentric Video Generation via 3D Full-Body Poses
57:01 Saurabh Gupta – Keynote Video Understanding for Robot Learning
1:27:37 Challenge Results
2:10:47 C V Jawahar – Keynote
2:43:37 EgoVis Distinguished Papers Award
3:09:02 Lorenzo Torresani – Where Perception Isn’t Enough: Anticipation, Guidance, and Reasoning Over Video
3:41:01 Willy Chung – Action 100M
3:46:58 Hao Li – Xperience 10M
3:51:43 Tengda Han – Unique Lives, Shared World: Learning from Single-Life Videos
3:59:43 Alexandros Deltizas – Reconstructing Functional 3D Scenes from Egocentric Interaction Videos
4:05:05 Hazel Doughty – The Devil is in the Details: Towards Fine-Grained Understanding with Limited Supervision
4:35:06 Ziwei Liu – From Egocentric Experience to Interactive Intelligence
5:04:55 CV Jawahar, Dima Damen, Saurabh Gupta, Hazel Doughty, Jakob Engel, Ziwei Liu – Panel Discussion
5:36:27 James Fort – Project Aria Updates
🔗 Full details, challenge leaderboards, and accepted papers: https://egovis.github.io/cvpr26/
Like and subscribe to the Meta Open Source Channel for more updates from CVPR 2026!
#CVPR2026 #CVPR #EgoVis #EgocentricVision #AI #ComputerVision #Robotics