Aug 2026· arXiv.org· Vol abs/2608.22888· 0 citations· 36 references
Computer Science
TL;DR
NemoSplat is presented, the first feed-forward 4D Gaussian Splatting framework tailored for media-aware dynamic reconstruction directly from uncalibrated marine videos, and a Media-Aware Gaussian Predictor is formulated to jointly estimate intrinsic 3D Gaussian attributes alongside physical media parameters, rendering pristine scene appearance in a single forward pass.
Abstract
Reconstructing photorealistic scenes in unconstrained underwater environments remains challenging due to severe media-induced light scattering and unpredictable dynamic objects. Recent feed-forward visual foundation models have demonstrated remarkable capabilities in generalized novel view synthesis and tracking. However, when directly applied to aquatic videos, optical attenuation and motion interference fatally corrupt their feature aggregation, leading to severe tracking and reconstruction failures. To overcome these limitations, we present NemoSplat, the first feed-forward 4D Gaussian Splatting framework tailored for media-aware dynamic reconstruction directly from uncalibrated marine videos. Beyond providing robust estimations of camera poses and dense scene depth, we devise a Promptable Dynamic Disentangler that utilizes a confidence-aware fusion strategy of learned dynamic probabilities and optional semantic text priors, effectively isolating massive transient entities. Furthermore, to counteract visual degradation, a Media-Aware Gaussian Predictor is formulated to jointly estimate intrinsic 3D Gaussian attributes alongside physical media parameters, rendering pristine scene appearance in a single forward pass. Additionally, we introduce a large-scale underwater dataset with massive dynamic elements to facilitate training and evaluation. Extensive experiments on our dataset demonstrate that NemoSplat achieves state-of-the-art tracking accuracy and high-fidelity rendering. Homepage: https://nemosplat.hkustvgd.com
Underwater 3D reconstruction remains challenging under sparse views, where scattering, absorption, and suspended particles degrade feature correspondences and geometric estimation. Although feed-forward geometry foundation models offer an alternative to conventional Structure-from-Motion, their direct application under...
H. Caldeira, Hao-Ran Wang, Guo-Xi Huang et al.· arXiv.org· 0 citations
A primary challenge in deep-sea 3D reconstruction is the strong coupling between active illumination and the participating water medium. Existing underwater 3D Gaussian Splatting (3DGS) methods typically model backscattering along pixel rays as a fixed constant, neglecting the distance-dependent attenuation of active l...
Underwater video enhancement (UVE) is essential for marine exploration, yet non-rigid refractive distortions and frequency-dependent attenuation often compromise the efficacy of standard restoration models. Although State Space Models (SSMs) like Mamba provide computational efficiency, their traditional direct-scan par...
Yue-Yue Wang, Wei-Ling Cai, Zhen-Xin Lin· Proceedings of the 34th ACM...· 0 citations
Prized for its high-quality real-time rendering, 3D Gaussian Splatting is a key novel view synthesis technology for fields such as virtual reality and autonomous driving, where it is widely used for outdoor scene reconstruction. However, extensive unbounded and texture-scarce regions in outdoor scenes, such as the sky,...
Jian Xiong, Qing Zhang, Jiu-Cheng Xie et al.· 0 citations
Underwater 3D reconstruction is critical for marine exploration, ecological monitoring, and subsea infrastructure inspection, yet remains challenging at large scale due to light attenuation, scattering, and limited capture coverage. While 3D Gaussian Splatting (3DGS) enables high-quality real-time rendering, its applic...
Hao-Ran Wang, Shao-Yu Cai, Adrian Azzarelli et al.· arXiv.org· 0 citations
We introduce HumetricNeRF, a depth-supervised dynamic neural radiance field (NeRF) for reconstructing humans in motion from sparse views. Our method leverages physical depth sensors, which provide reliable, albeit noisy, metric measurements and serve as supervisory signals for scene geometry reconstruction. When active...
Ori Francos, D. Raviv· IEEE Sensors Journal· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.