At CVPR 2026 in Denver, attendees observed computer vision research moving beyond isolated perception tasks toward integration with robotics and systems that simulate future states, support planning, and enable embodied interaction.
#CVPR'26 was a blast -- so much fun!
Lessons learned: - 3D is dead, long live 3D - no best papers for academia - vision is now robotics - AI generated papers vs AI reviews - everything is a world model - startup parties >> big tech - academia still no GPUs
Official programming included multiple events on video world models for robotics, active sensing, and closed-loop planning, matching the reported rise in accepted papers on embodied AI.
Attendee notes flag ongoing limits on GPU access for university labs and questions around AI-generated submissions, though these observations lack direct confirmation in the official program.
Many users find the shift from vision papers to robotics and world models exciting and agree startup parties outperform big tech, while others call related academic trends depressing because of AI-generated papers and real-world robot gaps.
@MattNiessner Pretty spot on @MattNiessner
#CVPR'26 was a blast -- so much fun!
Lessons learned: - 3D is dead, long live 3D - no best papers for academia - vision is now robotics - AI generated papers vs AI reviews - everything is a world model - startup parties >> big tech - academia still no GPUs
@MattNiessner Not sure if that was the intention, but that sounds ... pretty depressing
@MattNiessner So true - startup parties >> big tech
@MattNiessner Vision papers that crush test footage get humbled by mud, rust, weird angles.
Real robots live in that gap.
That's the actual story behind 'vision is now robotics.'
@Ferbin08 For robotics that makes of course sense, and of course I love that direction, but there's also many non-robotics use cases in vision.
@pretendsmarts Nope was not the intention - it was actually quite exciting :)
@MattNiessner At least for those 3, let's hope things will not stay that way for long: - no best papers for academia - AI generated papers vs AI reviews - academia still no GPUs
@MattNiessner The more things change the more they remain the same.
@Kuntalthakur12 🍾🍺🍻 Yup :D
@MattNiessner I agree with most … but why 3D is dead ? Infact 3D has alot of usage in robotics no ? Intresting findings btw
@MattNiessner Wow, unfortunately, all true and I noticed the same.. Why is 3D dead though?
@MattNiessner Can you talk more about vision is now robotics?
@MakkarNik @MattNiessner Summary
@MattNiessner wym by 3d is dead? everyone is working on 3d/4d for video understanding
@MattNiessner Highly agree on "startup parties >> big tech"
@MattNiessner I see a lot of VLAs papers but not world models...
@MattNiessner Sorry, I can fix this situation, but it still requires time. I spent 1.5 years on one answer and haven't finished yet.
At CVPR 2026 in Denver, attendees observed computer vision research moving beyond isolated perception tasks toward integration with robotics and systems that simulate future states, support planning, and enable embodied interaction.
#CVPR'26 was a blast -- so much fun!
Lessons learned: - 3D is dead, long live 3D - no best papers for academia - vision is now robotics - AI generated papers vs AI reviews - everything is a world model - startup parties >> big tech - academia still no GPUs