Gaussian splatting at Pix4D: a new era of 3D visualization
Gaussian splatting is making waves in the world of 3D mapping and visualization, promising a significant leap forward in the quality of point clouds and 3D reconstruction outputs. As one of the latest features available in PIX4Dcloud for users capturing with PIX4Dcatch, it opens up new possibilities for professionals across various industries. Here is everything you need to know about this game-changing feature, its capabilities, and the considerations to keep in mind as it evolves.
What is Gaussian splatting?
Gaussian splatting is an advanced processing technique that dramatically improves visualization in 3D models and generates more complete point clouds, producing a highly photorealistic representation of a scene. It generates outputs that are smoother, more detailed, and visually impactful.
Running Gaussian splatting doesn’t just enhance visualization—it also improves the underlying point cloud and 3D mesh. By refining how data is structured, Gaussian splatting ensures better results across all reconstruction outputs. This improvement makes it an excellent tool for all industries seeking detailed and realistic 3D reconstructions.
Gaussian splatting is a way to turn real-world scenes into highly detailed 3D models. Instead of using traditional sharp-edged shapes, it builds the model out of soft, semi-transparent “splats” (Gaussians) that blend smoothly to recreate surfaces. Imagine painting a picture by carefully layering tiny dots of color until the image looks just right—Gaussian splatting does something similar but in 3D. The result? A fast, realistic 3D model that’s great for visualization.
How it works
PIX4Dcatch introduces Gaussian splatting, a groundbreaking technical innovation that combines precise data capture with high-quality visualizations—all from your smartphone. Capture using PIX4Dcatch using an iPhone Pro or iPad Pro equipped with a LiDAR camera and seamlessly process on PIX4Dcloud. With this highly detailed reconstruction, you can understand your projects more clearly and identify exactly what needs further analysis.
PIX4Dcatch ensures accurate data capture even for large-scale projects, with support for datasets ranging from 50 to 2,000 images—both with and without RTK (Real-Time Kinematic) technology.
To achieve the best results with Gaussian splatting, proper scene calibration is key. Use the optimal image overlap, walk at a normal pace, and walk around your object of interest by keeping your phone in the same plane. If a scene has challenging surfaces with few features or large viewpoint changes, you can improve calibration by adding Pix4D Autotags in PIX4Dcatch. For general tips on achieving a well-calibrated scene, refer to the Pix4D documentation.
Settings can be adjusted per project and if you want to enable Gaussian splatting for older captures, simply reprocess them with Gaussian splatting enabled in your project settings. For more details on how to use PIX4Dcatch to generate a Gaussian splatting model, check out our technical documentation.
Can Pix4D's Gaussian splatting feature be georeferenced?
By default, Gaussian splatting operates in a relative coordinate system, meaning its positioning is based on the input dataset rather than absolute real-world coordinates. However, with PIX4Dcatch we can georeference the Gaussian splatting output in a projected coordinate reference system.
These Gaussian splats are fully georeferenced, allowing them to be visualized alongside other geospatial assets like point clouds and 3D meshes in the PIX4Dcloud viewer. By integrating high-quality visualization with precise georeferencing, Pix4D enables professionals to seamlessly analyze and compare different data layers in an interactive, accurate 3D environment.
Key benefits
- Unmatched visual quality: Gaussian splatting creates more complete point clouds which leads to more complete projects with exceptional detail. This is particularly valuable for professionals who rely on visually compelling outputs to communicate complex projects.
- Seamless workflow: the feature integrates effortlessly into the Pix4D ecosystem. Users can capture data with PIX4Dcatch, process it in PIX4Dcloud, and enjoy the improved outputs without additional complexity.
- Accuracy maintained: as with all Pix4D solutions, Gaussian splatting upholds the brand’s commitment to accuracy. Pix4D Autotags remain supported, ensuring precise results for users working on projects that demand technical excellence.
- Scalable processing: Gaussian splatting has been tested across projects of varying sizes, from 50 to 2,000 images. Processing times and costs scale accordingly, offering flexibility for projects of all complexities.
- Easy sharing: Gaussian splatting benefits from PIX4Dcloud’s easy-to-share feature, allowing users to quickly share their model via a simple URL, enabling seamless collaboration.
Who will be using this new feature?
Gaussian splatting is designed for professionals who require both accuracy and high-quality visualization. For example, surveyors can achieve more complete point clouds, by improving terrain mapping, site analysis, and as-built documentation. Utility professionals benefit from detailed 3D models that make it easier to visualize complex infrastructure with clarity, aiding in maintenance, planning, and decision-making processes. It’s a powerful solution for any industry that relies on highly detailed, realistic 3D models to enhance their workflows and insights.
Pix4D’s unwavering commitment to accuracy ensures that all outputs maintain precise data integrity while delivering enhanced visual clarity. The combination of technical excellence, impactful visualization, and the simplicity of capturing data with a smartphone makes Gaussian splatting a transformative tool. It is adaptable to a variety of use cases where clarity and detail are paramount.
Stay tuned for updates as Gaussian splatting continues to evolve, bringing even more possibilities to PIX4Dcatch and PIX4Dcloud users
