Gaussian splatting adjusts model size to improve 3D scene quality
Gauss What You Need: Compact Gaussian Splatting Across Scene Scales
Computer Vision and Pattern Recognition
Summary
3D Gaussian Splatting rebuilds a scene using many small shapes called Gaussians, but deciding how many to use is tricky and affects detail and speed. The authors found that the size of the area to capture a scene is known before training, while the detail level shows up only during training. They created TangoGS, a method that starts with a size guess based on the scene and then adjusts the number of Gaussians during training by checking image quality. This approach works well on both small and large scenes without needing to change settings a lot.
What this means in practice
- •For 3d graphics developers: Automatically scale 3D Gaussian models during scene reconstruction for better detail and rendering balance across small and large captures.
- •For augmented reality engineers: Improve AR scene models by dynamically adjusting Gaussian primitives to maintain quality without increasing computation unnecessarily.
Authors
Afif Boudaoud, Jiayi Liu, Alexandru Calotoiu, Torsten Hoefler
Abstract
3D Gaussian Splatting reconstructs a scene as a collection of Gaussian primitives from a set of posed photographs called the capture. The number of primitives used to represent the scene affects reconstruction quality, storage, and rendering cost. How to select this number automatically across capture scales remains unresolved: configurations effective on standard benchmarks can leave larger captures with too few Gaussians to reconstruct fine details. We observe that the surface to represent, given by the capture's extent and resolution, is known before training, whereas its content complexity becomes apparent during training, through the reconstruction quality on the training views. We introduce TangoGS, which combines capture-derived model sizing with training-based adaptation: the capture determines the scale of the model, and training feedback determines its final size within that scale. Before training, TangoGS derives a learning allowance for model growth from the capture's total pixels after discounting views that re-observe the same scene points. During training, reconstruction quality guides how many Gaussians to add and remove. On 13 standard benchmark scenes, TangoGS matches the mean PSNR of the best-performing evaluated baseline, LeGS, with $48\%$ fewer Gaussians. On eight large captures, the same configuration automatically scales to larger models when necessary, achieving the highest mean PSNR among evaluated methods: $0.54$ dB above the runner-up with $2.3\times$ as many Gaussians. Together, capture-derived learning allowances and training-quality guided density control enable a state-of-the-art quality--size compromise across scene scales without retuning.