Computer Vision News - August 2022

18 AI Application required for a portrait mode effect and to blur the background. They were very resistant to putting in more cameras. The executive eventually saw the benefits, but making that change was an uphill battle. ” Tom tells us he enjoyed attending the Computational Cameras and Displays workshop at CVPR this year. Metalenses was a hot topic, which he believes could complement Glass’s work in time. “ The idea of using wave optics designs to make very thin lenses has been shown to have some great potential and is used in some research environments, ” he says. “ Therearehopes toput that intocommercial photography as well, but there are some practical constraints, such as scaling up the systems to large sensors and working across multiple wavelengths. That’s going to be technology that takes longer to get ago, a great attempt to take portrait mode into video, but it still suffers from those artifacts around the edges. We can do it naturally, just like in a Hollywood movie. You can shoot that on a smartphone with our technology! ” Glass is applying its knowledge of the whole manufacturing and supply chain, design, optics, sensing, algorithms, and what can be run on a smartphone chip. Tom, Ziv, and the team have accumulated much interdisciplinary expertise across those areas throughout their careers. “ That enables us to look at other trade-offs and design constraints that we can open up, ” Tom points out. “ Back at Apple, we had to push hard to go to a dual camera when single cameras were still being used in smartphones. We needed the stereo information to create the depth maps Render showing how a Glass camera module integrates into a smartphone - note the large entrance window