Computer Vision News - November 2022

15 Ishit Mehta many papers are coming out every week, and trying to keep up with that literature while writing papers and doing relevant work is a challenge. You might have solved something, or you might have an interesting idea, but if you are six months late, then the relevancy of that idea goes down. ” The work brings ideas from many areas, including geometry processing , computer vision , and computer graphics . These areas all intersect but operate in silos. People working in these specific areas do not always read the literature from other areas. The challenge here is how to write things that are relevant to all three communities simultaneously and are interesting and useful. “ The inception of the idea behind the paper comes from the inverse rendering community , where you’re given a set of images in a constrained environment or an unconstrained environment, and you want to recover all the properties of the scene where those images are captured, ” Ishit explains. “ For instance, you have a glass, external constraints and want to recover what the shape looks like. Say you are designing an aircraft, and you know what the plane is going to fly, what the air is going to look like, and what speed it’s going to fly at, and you want to design it so there is minimal drag and it runs efficiently with the least amount of fuel consumption. You may want to design a bridge to have maximum strength using the least amount of material, and you know how many people will walk on it, the force field and pressures of the river, and the type of soil involved. Maybe, using computational fluid dynamics , you are designing efficient aerodynamic components for high-speed Formula One race cars. In all these kinds of settings, geometry optimization is essential. Did Ishit encounter any challenges in the course of this work? “ Oneof thebiggest problemsworking in this space is there is just too much happening, ” he points out. “ This is a problem faced by most PhD students working in this area. So