Making the Most of Your Camera | James Tompkin

Abstract: Images are everywhere, especially images of the real world, and visual computing is important both for reconstructing useful models from these images and for providing us humans with interactive tools for visualization and analysis. These methods aid real world sensing and measurement, scientific and medical imaging, and media and the arts. The past three years in visual computing have been populated by methods in neural fields – a flexible way to solve the inverse problems required to reconstruct visual scene models. To make the most of the many images from our cameras, we must be able to scalably reconstruct large scenes from thousands of images, and I will discuss how to achieve this with hybrid neural fields. Further, to make the most of active illumination sensors in our cameras, we must be able to integrate their different signals, and I will discuss a physically-based neural field to achieve this, including for dynamic scenes. Finally, I will contextualize these tools within ongoing discussions around data and 3D learning.

Bio: James Tompkin (jamestompkin.com) is the John E. Savage Assistant Professor of Computer Science at Brown University. His research at the intersection of computer vision, computer graphics, and human-computer interaction helps develop new visual computing tools. His doctoral work at University College London studied large-scale video processing and exploration techniques, and postdoctoral work at Max-Planck-Institute for Informatics and Harvard University helped create new methods to edit content within images and videos. Recent research has developed new techniques for low-level scene reconstruction, view synthesis for VR, and content editing and generation.