SuperSplat Hosts High-Resolution 3D Gaussian Splat Model of Strawberry

A 3D Gaussian Splat model of a strawberry has been published on the SuperSplat platform, offering a detailed digital representation of the fruit. The model was generated using 88 focus-stacked images captured from 90 distinct perspectives. This approach leverages Gaussian splatting, a technique designed for high-fidelity 3D scene reconstruction and rendering, allowing users to interact with the digital object in a spatial environment.

The imagery was captured using a Nikon Z8 full-frame camera equipped with a Laowa 180mm macro lens. The photographic setup utilised LED lighting to ensure consistent illumination across the capture sequence. Specific camera settings included an aperture of f/7.1, an exposure time of 1/160, and an ISO of 100, parameters selected to optimise depth and clarity for the subsequent 3D processing.

Focus stacking was employed as a core component of the image capture strategy. This photographic technique increases the depth of field in an image by combining multiple images taken at different focus distances. By integrating these stacked images, the resulting dataset provides the necessary visual data for the Gaussian splatting algorithm to construct a coherent three-dimensional model.

SuperSplat serves as the designated platform for viewing and interacting with these 3D Gaussian Splat models. The publication of the strawberry model on the platform highlights the accessibility of such reconstruction tools, which are increasingly relevant in fields requiring precise digital twins of physical objects. The model is available for exploration via the SuperSplat project page.

While the technical specifications of the capture process are clearly documented, the source material does not provide the specific date of publication beyond the feed timestamp, nor does it identify the photographer or creator of the model. The exact lighting configuration, including the number of LEDs or colour temperature, is also not fully detailed in the available records. The announcement was initially surfaced through the Hacker News feed, directing users to the primary project page for further metadata.