Netsuke 3D

When I was working at the Perceptual Robotics Labs of Scuola Sant’Anna, we were approached by the local museum of graphics to help with an exhibition they were planning. They wanted to put on show the private collection of Netsuke from Prof. Edda Brescianini, a world-renowed archaeologist and Egyptologist, and needed us to come up with a way to enhance the visualization of these small artifacts.

A small figurine of two human-like figures. The larger figure is standing and holding a staff, while the smaller figure is kneeling and holding an object resembling a bowl or fruit. The figurine appears to be made of a light-colored material, possibly ivory or ceramic, and is placed on a dark circular platform. The background is blurred, emphasizing the figurine itself.

For those of you who are not familiar with Netsuke, they are tiny Japanese sculptures, often made of wood or ivory, that were used as buttons on Japanese traditional clothing between the XVII and XX centuries. While being rich in details, Netsuke are extremely small, often below 2cm on their longest side, and it may not be easy to have a closer look while visiting the museum.

After some team discussions, we decided to opt for the simplest, yet most effective solution: we would recreate a 3D model of a few of these Netsuke and display them on a big TV near the real objects. The museum sees a lot of visitors during these exhibitions and any sort of VR experience would’ve been hard to handle.

A statue of a person kneeling on a platform in a Zen garden. The garden features concentric circles raked into the sand, and there is a leafless tree in the background. In the top left corner, there is Japanese text written in red, which reads '鯉釣り (Koi Tsuri),' translating to 'Carp Fishing

Getting the 3D model of a Netsuke was not an easy task: they are extremely small and very shiny, and despite being very strong we did not feel comfortable trying any invasive technique. Laser scanning was our first option, but we would lose a lot of colors, and it was therefore discarded. The only viable solution was to use photogrammetry, but it’s not so easy when the object is this small. We tried different approaches, such as using a macro lens and a soft-box, but they all failed for one reason or another. The final model was obtained by using a longer focal length from distance during a cloudy day, outdoor, using a light diffuser to prevent even the smallest reflection on the object, and with a tiny bit of sculpting in post processing to get rid of lens distortion.