ParetoGS: High-Quality Continuous-Budget 3D Gaussian Splatting with Sparse Adapted Anchors
ParetoGS is a method for scalable 3D Gaussian Splatting that serves many Gaussian-count budgets from one representation. It builds a small set of adapted budget anchors, stores their parameter changes in a compact residual atlas, and interpolates between anchors to decode intermediate budgets without separate optimization for each. Across 13 scenes, it reports better average PSNR, SSIM, and LPIPS than compared prefix methods at 10% of each method's full model, and it cuts mean additional state from 390.0 to 85.7 MB with a 0.01 dB average PSNR change.