Hello, World!
Welcome to the Holoscanner blog! Here, we’ll post regular updates on the development of our Hololens app. This project is part of the first ever AR/VR Capstone at the University of Washington, where teams of four students get access to multiple Hololens devices to make a holographic app in a single 10-week quarter.
The Story
Our team of PhD students is in the interesting position of having the deep technical knowledge, early access to the Hololens hardware, and direct collaboration with the Hololens team at Microsoft. We wanted to work on an application that had interesting, research-like challenges that we were uniquely qualified to solve, rather than just another AR game that any developer with a Hololens could implement.
Several of us do research related to room-scale 3D scanning and reconstruction, and we commonly encounter problems with incomplete or inaccurate scans. Getting good scans is an art. The sophisticated tracking and reconstruction capabilities of the Hololens make it a promising scanning device, and its user interface opens up an interesting questions: if everyone has a Hololens, can everyone get expert-quality 3D scans? Holoscanner is our first stab at enabling this vision.
The Plan
Our app will have two main parts: the frontend running on the Hololens, with an engaging game, and the backend running in the cloud, which will refine the scans and analyze them to determine what regions need further scanning.
The game will be based around the mechanic of locating objectives in the scene, which will be hidden in the regions that need more scanning. We start with the barebones treasure-hunting interpretation of this mechanic, but might extend it if time permits.
The backend involves the two interesting research problems of mesh refinement and region-quality evaluation. Combining the 3D data from multiple devices and scans is a fairly well-studied problem, but tweaking existing methods for the unique demands of augmented reality applications might be challenging, especially when combined with latency and bandwidth concerns. Meanwhile, identifying inconsistent, incomplete, or difficult-to-scan regions is a fairly novel problem and will be very interesting to investigate.
More details can be found in the Proposal and Project Requirements Document.
Edward Zhang