Our team at OAO set out to develop a 3D scanning technology for GIMII that would be fast, high-quality, and accessible. However, achieving this goal was no easy feat. We faced the challenge of creating an application that could be used by medical professors and students with little to no experience in 3D scanning while still being able to capture highly detailed models, including even the smallest features like a needle.
Our first approach to solving the 3D scanning problem is to use a technique called photogrammetry. This well-developed technology is used in many areas, such as game development and architecture, and there are software solutions like Reality Capture and Object Capture that are excellent for scanning high-quality assets. However, while these solutions are well-suited for many use cases, they don't quite fit the specific needs of GIMII, such as scanning dissection specimens in real time. Unlike other applications, these models cannot be re-scanned during dissection, so users need immediate feedback on which angle to scan from, leaving no room for error.
With these limitations in mind, we began developing our own solution for 3D scanning. Thanks to recent advancements in AR technology, both Apple and Google now provide great API support for developing AR applications. We have decided to create an AR-enabled 3D scanning application that provides real-time guidance to users to help them capture high-fidelity models, even the smallest details. Our goal is to launch this application by the end of the summer as part of a limited beta release alongside GIMII, allowing us to fully test the capabilities of our technology.
Photogrammetry is a great technique for creating 3D models, but there is always room for improvement. One challenge we face when scanning specimens is dealing with reflective surfaces. To address this issue, we are actively exploring new techniques that can capture 3D models with even higher fidelity. One exciting development we are currently working on is NeRF (Neural Radiance Field), which leverages AI to capture models that can even reproduce reflection and refraction. Although NeRF is still in the early stages of development, we are eagerly anticipating the future of this technology and the possibilities it could bring to 3D scanning.