Abstract: Human intelligence is beyond pattern recognition. From a single image, we’re able to explain what we see, reconstruct the scene in 3D, predict what’s going to happen, and plan our actions accordingly. In this talk, I will present our recent work on physical scene understanding—building versatile, data-efficient, and generalizable machines that learn to see, reason about, and interact with the physical world. The core idea is to exploit the generic, causal structure behind the world, including knowledge from computer graphics, physics, and language, in the form of approximate simulation engines, and to integrate them with deep learning. Here, deep learning plays two major roles: first, it learns to invert simulation engines for efficient inference; second, it learns to augment simulation engines for constructing powerful forward models. I’ll focus on a few topics to demonstrate this idea: building scene representation for both object geometry and physics; learning expressive dynamics models for planning and control; perception and reasoning beyond vision.
MORE ABOUT THE SPEAKER:
Jiajun Wu is an Assistant Professor of Computer Science at Stanford University, working on computer vision, machine learning, and computational cognitive science.
How to Attend
Registration is requested. Use the Sign Up link below.Sign Up
Organized By: Stanford HAI