CAdam: Context-Adaptive Moment Estimation for 3D Gaussian Densification in Generative Distillation

2026-05-20

AI Summary

This work addresses the challenge in existing 3D Gaussian densification for generative distillation, where random guidance often yields redundant primitives, struggling to balance over-densification and under-fitting. The authors reformulate the problem as a statistical signal validation task and propose CAdam, a novel framework that introduces an interference mechanism inspired by signal processing. By leveraging the first-order moment of gradients to disentangle geometric signals from generative noise, and integrating quantile-based context awareness with signal-to-noise ratio (SNR) gating, CAdam enables adaptive densification with soft termination. Evaluated across diverse generative targets and backbone architectures, the method reduces Gaussian counts by 85%–97% while preserving comparable perceptual quality, substantially improving memory efficiency.

Abstract

Adaptive densification is the engine of 3D Gaussian Splatting (3DGS). However, when transposed to the optimization-based Generative Distillation paradigm, this reconstruction-native mechanism reveals fundamental limitations, resulting in inefficient representations cluttered with redundant primitives. We diagnose this failure as a Densification Dilemma stemming from the stochastic nature of generative guidance: the standard magnitude-based accumulation indiscriminately aggregates transient noise alongside geometric signals, making it difficult to strike a balance between over-densification and under-fitting. To resolve this, we introduce Context-Adaptive Moment Estimation (CAdam), a novel framework that reinterprets densification as a statistically grounded signal verification problem. CAdam leverages the first moment of gradients to exploit the interference principle, where stochastic fluctuations cancel out via destructive interference while consistent geometric drifts accumulate via constructive interference, effectively disentangling the underlying signal from the generative noise floor. This is further augmented by a quantile-based context awareness and an intrinsic Signal-to-Noise Ratio (SNR) gating mechanism, which ensure robust adaptation across optimization stages and enable the soft termination of densification. Extensive experiments across diverse objectives (SDS, ISM, VFDS) and strong generative 3DGS backbones show that CAdam reduces Gaussian count by 85%-97% relative to standard densification while preserving overall comparable perceptual quality. These results highlight signal-aware density control as a practical way to improve memory efficiency in optimization-based generative distillation.

Problem

• 3D Gaussian Splatting
• Generative Distillation
• Adaptive Densification
• Signal-to-Noise Ratio
• Redundant Primitives

Innovation

• Context-Adaptive Moment Estimation
• 3D Gaussian Splatting
• Generative Distillation
• Signal-to-Noise Ratio gating
• adaptive densification