Pete Florence: Dense Visual Representations, NeRFs, and LLMs for Robotics
The Gradient: Perspectives on AI
00:00
How Does a Diffuse 3D Modeling Work in Computer Vision?
The idea of correspondences, whether they're learned or otherwise retrieved in computer vision is like, you know, there's this great quote by Takeo Kanad. He says the three fundamental problems in computer vision are correspondence, correspondence and correspondence. So as you mentioned, this can be trained via self supervised learning. And then my friend Lucas Manueli and I, we were working on this together in grad school and we came up with some ideas to make it work better.
Transcript
Play full episode
