Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces InceptionGS, a novel method that enhances Gaussian splatting for large-scale scene digitization by addressing the challenges posed by unstructured view sampling. By integrating scene- and view-adaptive generative priors, the method effectively repairs under-sampled regions while maintaining rendering quality in well-covered areas. Extensive experiments confirm that InceptionGS outperforms existing reconstruction and generation methods, achieving high-fidelity results even in complex environments with sparse observations.
InceptionGS redefines scene digitization by seamlessly blending reconstruction and generative techniques to tackle the pitfalls of unstructured view sampling.
Achieving truly immersive large-scale scene digitization necessitates consistent and visually pleasing rendering across all possible viewing perspectives. However, collecting multi-view images covering every fine detail of a large-scale scene is prohibitive due to scene complexity, capture cost, negligence, or accessibility constraints. As a result, the sampled views tend to be highly unstructured -- the majority of the scene is well covered yet certain regions inevitably lack sufficient observations. Existing reconstruction based methods are vulnerable to view scarcity while generation based approaches suffer from generalization, controllability, and 3D consistency issues. To address this challenge, we propose InceptionGS, which bootstraps Gaussian splatting by subtly balancing reconstruction and generation. Starting from an initial Gaussian splatting, InceptionGS reasonably rethinks and repairs problematic regions caused by view scarcity while preserving the quality elsewhere, by softly incorporating scene- and view-adaptive generative priors. Extensive experiments on real-world large-scale scenes demonstrate the superiority and broad applicability of our approach in handling unstructured imagery and boosting high-fidelity Gaussian splatting. Please refer to the supplementary video for better visual demonstrations.