Computer Vision News - June 2022

5 CCD with Emma Alexander Every advance in mainstream computer vision gets multiplied and exponentially expanded in terms of the potential once you let every single component change. We’re never going to run out of new computational cameras and displays. ” One area where the workshop offers broad appeal to the mainstream CVPR audience is self-driving cars. It is widely acknowledged that regular cameras do not cut it in this domain. This year, the event will be expanding its optical horizons by looking at LiDARs and SPADs . With the workshop schedule at CVPR typically full to capacity, what will make CCD stand out from the crowd? “ It’s going to be a very creative workshop – you’re going to see and think about things in new ways, ” Emma reveals. “ When you break out of the mold of what a camera or a display is, you can bring back the magic of science and problem- solving . I always learn, not just new facts or new techniques, but whole new perspectives on the field by finding out about these creative, innovative ways a camera where the optics and the algorithm are co-designed to make scene information available . That could be moving some of the computation into theoptics touse thephysics of light topre- process data and save on computational efficiency, or it could be using lots of computation to exceed the native limits of the optical system. Doing this unlocks enormous potential in both cameras and displays, where instead of using standard datasets and standard photographic and video data, the optical pipeline can be altered to create new kinds of data to access and display information. Ten years after its first outing, what keeps people coming back to CCD? “ There’s new work coming out every year and a great deal of space to explore in displays and cameras, ” Emma tells us. “ Standard cameras are very simple. It’s a lens and a sensor. But there are plenty of new designs and technologies even within the components that we know – for example, just arranging lenses and filters in new ways or pairing them with new algorithms. There’s so much to explore. Ulugbek Kamilov