Gradient-Driven Natural Selection for Compact 3D Gaussian Splatting
This work addresses the need for more efficient 3D scene representation in computer vision, offering a fully learnable method to reduce resource usage without manual intervention.
The paper tackles the problem of high storage and computational overhead in 3D Gaussian Splatting (3DGS) by proposing a natural selection-inspired pruning framework that autonomously determines which Gaussians to retain or prune, achieving over 0.6 dB PSNR gain under 15% budgets.
3DGS employs a large number of Gaussian primitives to fit scenes, resulting in substantial storage and computational overhead. Existing pruning methods rely on manually designed criteria or introduce additional learnable parameters, yielding suboptimal results. To address this, we propose an natural selection inspired pruning framework that models survival pressure as a regularization gradient field applied to opacity, allowing the optimization gradients--driven by the goal of maximizing rendering quality--to autonomously determine which Gaussians to retain or prune. This process is fully learnable and requires no human intervention. We further introduce an opacity decay technique with a finite opacity prior, which accelerates the selection process without compromising pruning effectiveness. Compared to 3DGS, our method achieves over 0.6 dB PSNR gain under 15\% budgets, establishing state-of-the-art performance for compact 3DGS. Project page https://xiaobin2001.github.io/GNS-web.