InceptionGS is a method that improves large‑scale Gaussian splatting for scenes captured with unstructured view sampling. It starts from an initial Gaussian splatting and then selectively repairs regions that suffer from sparse views by incorporating adaptive generative priors, while preserving quality in well‑sampled areas. Experiments on real‑world scenes show that this balanced reconstruction‑generation approach yields higher‑fidelity results and works broadly across unstructured imagery.
F4Splat introduces a feed‑forward predictive densification strategy for 3D Gaussian splatting that allocates Gaussians based on a densification‑score guided by spatial complexity and multi‑view overlap. The method predicts per‑region scores to estimate required Gaussian density, enabling explicit control over the total Gaussian budget without retraining. This adaptive allocation reduces redundancy in simple regions and minimizes duplicate Gaussians across overlapping views, yielding compact yet high‑quality 3D representations and superior novel‑view synthesis performance with fewer Gaussians.
By Injae Kim, Chaehyeon Kim, Minseong Bae, Minseok Joo, Hyunwoo J. Kim
The paper introduces a fusion‑aware hierarchical Gaussian patch representation that enables direct class‑guided generation of 3D Gaussian Splatting (3DGS) objects. By decomposing irregular Gaussian sets into canonical local patches and encoding them as structured tokens, the method fuses global class semantics with patch‑level geometry, appearance, spatial correspondence, and rendering‑sensitive cues. A structure‑aware rectified flow model, conditioned on patch positions and coupled with global‑local velocity prediction and density‑aware weighting, produces class‑conditioned 3DGS objects within seconds, achieving more coherent geometry, sharper local details, and better multi‑view consistency than baseline models.
By Yizhao Wang, Jingbo Wang, Guantao Zhang
InceptionGS is a method that improves large‑scale Gaussian splatting for scenes captured with unstructured view sampling. It starts from an initial Gaussian splatting and selectively repairs areas affected by sparse views by integrating scene‑ and view‑adaptive generative priors, while keeping well‑covered regions unchanged. Experiments on real‑world scenes show that this hybrid reconstruction‑generation approach yields higher‑fidelity results than existing methods.
By Tianheng Lu, Guangyu Wang, Ruqi Huang, Lu Fang
Spackle is a lightweight residual learning framework designed to improve large-view single-image novel view synthesis (NVS) by mitigating capacity competition in hybrid decoupled systems that combine 3D Gaussian Splatting (3DGS) and diffusion models. It operates in three stages: predicting base 3DGS attributes, automatically identifying poorly reconstructed regions, and learning a residual 3DGS focused on those areas. During inference, Spackle merges the baseline and augmented Gaussians to produce high-fidelity novel views, achieving state‑of‑the‑art performance on large-view-deviation cases.
By Xuanzhi Liu, Yuhe Zhou, Xinyi Wu, Zhenyao Wu, Jinghao Chen, Ruize Han, Song Wang
KISS-GS is a modular compression pipeline for 3D Gaussian Splatting (3DGS) scenes that separates compression from training. It first compacts a vanilla 3DGS scene by 15.7× using state‑of‑the‑art pruning, then encodes the result into the SOG‑XT image‑based format, achieving an additional 6.6× reduction. Optional encoding‑aware fine‑tuning can further cut the size by 2.2×, yielding total reductions of 85× to 319× on standard benchmarks while enabling web‑native decoding.
By Wieland Morgenstern, Friedrich Elias Branschke, Florian Fleischmann, Adrian Szatmari, Paul Schlack, Florian Barthel, Peter Eisert, Anna Hilsmann
Bi-FlowGS introduces a bidirectional co-refinement framework that links generative view completion with 3D Gaussian Splatting geometry. It employs Video-to-Geometry Flow Distillation (V2G) to transfer temporal correspondence from restored videos into Gaussian geometry, mitigating the Geometry Cheating problem. Simultaneously, Geometry-to-Video Flow-Guided Restoration (G2V) uses the current 3DGS geometry to guide temporally consistent video restoration, creating a loop where restored videos and optimized geometry iteratively improve each other, leading to better rendering quality and geometric consistency on wide-baseline and 360° benchmarks.
By Yuetong Wang, Jinsheng Quan, Yi Yang, Yawei Luo
arXiv:2609.22941v1 Announce Type: new
Abstract: Novel view synthesis from sparse inputs remains challenging for 3D Gaussian Splatting (3DGS) due to ambiguous geometry, cross-view inconsistency, and m...
By Yunqi Gao, Zhanfeng Liao, Hanzhang Tu, Zhaoqi Su, Guoqing Zheng, Songtao Wang, Hongwen Zhang, Zhou Xue, Leyuan Liu, Yebin Liu
arXiv:2607.20813v2 Announce Type: replace
Abstract: Pixel-aligned Gaussian splatting enables efficient and generalizable novel-view synthesis. However, high-resolution rendering faces a critical trad...
By Jiun Lee, Jaekwang Kim, Sangmin Lee
arXiv:2508.03077v2 Announce Type: replace
Abstract: Feedforward 3D Gaussian Splatting (3DGS) overcomes the limitations of optimization-based 3DGS by enabling fast and high-quality reconstruction with...
By Anran Wu, Long Peng, Xin Di, Xueyuan Dai, Chen Wu, Yang Wang, Xueyang Fu, Yang Cao, Zheng-Jun Zha
The paper introduces TruncGradGS, a piecewise truncated gradient approach that mitigates gradient vanishing in 3D Gaussian Splatting, enhancing optimization stability and robustness to initializations. It demonstrates consistent improvements over random and COLMAP initializations in both static and dynamic settings. Additionally, the authors highlight limitations of existing dynamic scene benchmarks and present a new synthetic dataset for evaluating dynamic Gaussian Splatting.
By Theo Morales, Nhat-Quynh Le-Pham, Robin Atkins, Binh-Son Hua
VoxelTTO is a feed‑forward framework that reconstructs 3D Gaussian splatting scenes from multiple images by aggregating dense image features into a global voxel representation and decoding Gaussians from voxel features, thereby eliminating the pixel‑to‑Gaussian correspondence. It incorporates test‑time optimization with lightweight LoRA modules to adapt to known camera parameters while keeping the pretrained visual foundation model frozen. The method replaces standard rasterization with stochastic solid volume rendering, improving geometric fidelity, and demonstrates superior RGB‑D novel‑view synthesis and camera‑pose estimation on Replica, Tanks and Temples, and DTU datasets.
By Yibin Zhao, Yihan Pan, Yangwen Li, Jun Nan, Jianjun Yi