Convert 3D Gaussian Splatting to 3DTiles Using GISBox

With the continuous development of 3D visualization technology, 3D Gaussian Splatting, as an efficient 3D rendering technology, has been widely used in digital twins, smart cities and other fields. So why do we need to convert 3DGS files to 3DTiles? What tools should we use to convert? This article will introduce you to the appropriate GIS tools and specific conversion steps.

I. Specific operation steps
(1) Task creation and file addition

1. Start GISBox: Open the GISBox software and make sure that the required environment has been correctly installed and configured.

2. Create a new slicing task: In the toolbar on the left side of the software, click the "Slice" function, and then click the "+ New" button in the upper right corner to create a new slicing task.

3.Add Gaussian splatting file: In the opened slicing task window, select the "Gaussian splatting" type. Next, click the "Add File" button and select the 3D Gaussian Splatting file you need to convert from the local file system. GISBox will use its powerful data parsing capabilities to quickly read the file and build the corresponding 3D model data structure. If the file data volume is large, such as containing city-level point cloud data, it is recommended to check the integrity of the file in advance to avoid errors during the conversion process.

(2) Parameter configuration and conversion execution

1. Parameter setting: In the slicing task window, you need to configure the conversion parameters. This includes: Export path: Select the path to save the converted 3DTiles file. Spatial reference: Select the appropriate spatial reference system according to the actual data situation. Zero point coordinates: Set the zero point coordinates of the data model to ensure that there is no offset during rendering.

1. Execute the conversion: After confirming that all parameter settings are correct, click the "OK" button. GISBox will start the conversion process according to the set parameters, and you can view the progress and status of the conversion in real time on the software interface.

(3) Application and precautions after conversion

1. Model service viewing: After slicing is completed, you can view the corresponding model service in the "Distribution" column of GISBox. These service addresses can be applied to GIS engines such as Cesium and Cesium for Unreal, and can also be integrated into digital twin project construction platforms such as Shanhaijing Visualization.

2. Data accuracy maintenance: When preprocessing and converting data, be sure to maintain data accuracy. Especially when simplifying the model, avoid distortion of the model geometry due to over-simplification, which will affect the final display effect.

3. File size optimization: The size of the converted 3DTiles file will directly affect the data loading speed. Therefore, before conversion, it is recommended to reduce unnecessary data redundancy and optimize the data structure through data preprocessing to reduce the size of the generated 3DTiles file and improve the data loading efficiency in network transmission and application.

4. Cesium compatibility: If you plan to use the converted 3DTiles file in GIS engines such as Cesium, please make sure that the Cesium version supports the KHR_gaussian_splatting extension. Because the slice output of the Gaussian splatter model requires this extension to render correctly. You can get the Cesium branch that supports this extension from GitHub, compile and load it.

III. Summary
The conversion from 3D Gaussian Splatting to 3DTiles through GISBox not only reduces the technical threshold, but also significantly improves the rendering efficiency of data on the Web side. Whether it is digital twins, smart cities, or cultural heritage protection, this process can provide strong support for the efficient rendering of three-dimensional scenes. I hope this article can be helpful to everyone.