Shivansh Rao of GLASS Imaging

This talk explores the evolving landscape of AI-powered computational photography and imaging, focusing on how deep learning is redefining traditional camera pipelines. We begin by examining the data foundation—from real-world RAW and RGB image collection to synthetic data generation strategies that expand training diversity and robustness. Building on this, we delve into the design of Neural Image Signal Processors (Neural ISPs)—learning-based pipelines that perform demosaicing, denoising, tone mapping, and color correction in a unified framework. The session highlights practical methods to make neural ISPs fast, controllable, and deployable, bridging the gap between research prototypes and production-ready systems. Attendees will gain hands-on insight into dataset design, simulation, and model evaluation, and leave with a deeper understanding of how AI can push the limits of image quality, performance, and creative control.

Biography:

Shivansh leverages expertise in computer vision, signal processing, and deep learning to develop image registration and restoration algorithms for mobile imaging. Prior to Glass, he worked at Qualcomm R&D on hardware-efficient algorithms for 6DoF head-pose-tracking and 3D Reconstruction. He earned his Master's degree from Penn State and Bachelor's from Delhi College of Engineering and has conducted research work at IIT-Kanpur, IIT-Delhi along with collaborative research with Google AI. His research is published at CVPR, ECCV, ACM-MM, and others, along with pending US patent applications.

Address:United States