aiqbench analysis
CGGS advances text-to-3D generation by integrating consistency-augmented diffusion models with geometric refinement to solve viewpoint alignment issues in ego-centric scenes. The method leverages optical flow and entropy-based depth losses within a 3D Gaussian Splatting framework to produce coherent, high-fidelity reconstructions from limited inputs. This work represents a significant step forward in generating semantically aligned and structurally accurate three-dimensional environments directly from textual descriptions.
CGGS: Consistency-Augmented Geometric Gaussian Splatting for Ego-Centric 3D Scene Generation
arXiv:2607.03819v3 Announce Type: replace-cross Abstract: Challenges remain in ego-centric 3D scene generation due to limited view overlap and the dominant influence of individual perspectives on scene interpretation. These factors hinder the creation of viewpoint-consistent and semantically aligned visual content, as well as the construction of accurate geometric structures. In this paper, we propose CGGS, a text-to-3D framework aiming to enhance 3D-content-awareness and address geometric distortions in ego-centric scene generation. Firstly, the Ego-centric Generator is proposed by fine-tuning a Multi-View Latent Diffusion Model with consistency-augmented loss to generate consistent, high-fidelity 2D content aligned with textual descriptions. Then, Layout Decorator leverages optical flow and point-track correspondence to estimate depth, therefore producing dense point clouds as coarse layouts from the ego-centric 2D priors. Building on this initialization, Geometric Refiner is proposed to enhance 3D Gaussian reconstruction via an entropy-based Mutual Information Depth Loss (MID) combined with a hierarchical optimization scheme for improving visual quality and geometric structure. Comprehensive experiments demonstrate that CGGS outperforms previous methods in generating coherent and accurate text-driven 3D scenes.
Abstract:Challenges remain in ego-centric 3D scene generation due to limited view overlap and the dominant influence of individual perspectives on scene interpretation. These factors hinder the creation of viewpoint-consistent and semantically aligned visual content, as well as the construction of accurate geometric structures. In this paper, we propose CGGS, a text-to-3D framework aiming to enhance 3D-content-awareness and address geometric distortions in ego-centric scene generation. Firstly, the Ego-centric Generator is proposed by fine-tuning a Multi-View Latent Diffusion Model with consistency-augmented loss to generate consistent, high-fidelity 2D content aligned with textual descriptions. Then, Layout Decorator leverages optical flow and point-track correspondence to estimate depth, therefore producing dense point clouds as coarse layouts from the ego-centric 2D priors. Building on this initialization, Geometric Refiner is proposed to enhance 3D Gaussian reconstruction via an entropy-based Mutual Information Depth Loss (MID) combined with a hierarchical optimization scheme for improving visual quality and geometric structure. Comprehensive experiments demonstrate that CGGS outperforms previous methods in generating coherent and accurate text-driven 3D scenes.
Graphics (cs.GR); Artificial Intelligence (cs.AI)
arXiv:2607.03819 [cs.GR]
(or arXiv:2607.03819v3 [cs.GR] for this version)
IEEE Transactions on Image Processing, vol. 35, pp. 7670-7684, 2026
https://doi.org/10.1109/TIP.2026.3710486 DOI(s) linking to related resources
Full text under the publisher's open licence · 194 words
Related coverage
CoRef-GS: Cooperative Referring Gaussian Splatting for Multi-Agent Scene Understanding
arXiv · Sep 18, 2026
OpenSplatGraph: From Dense Semantic Maps to Structured Scene Graphs for Open-Vocabulary Robot Perception
arXiv · Oct 7, 2026
VG3S: Visual Geometry Grounded Gaussian Splatting for Semantic Occupancy Prediction
arXiv · Jul 23, 2026
OctoSplat: Hybrid OctoMap-Gaussian Splatting for Active Semantic Mapping and Phenotyping with Horticultural Robots
arXiv · Aug 17, 2026