NemoSplat: Feed-Forward 4D Gaussian Splatting for Media-Aware Underwater Reconstruction

• URL: http://arxiv.org/abs/2608.22888v2
• Date: Tue, 25 Aug 2026 04:52:42 GMT
• Title: NemoSplat: Feed-Forward 4D Gaussian Splatting for Media-Aware Underwater Reconstruction
• Abstract summary: NemoSplat is the first feed-forward 4D Gaussian Splatting framework tailored for media-aware dynamic reconstruction directly from uncalibrated marine videos.<n>NemoSplat achieves state-of-the-art tracking accuracy and high-fidelity rendering.
• Score: 21.224361660534814
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• 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

Related papers

• 3D-USE: From Image-Level to Scene-Level Underwater Enhancement [54.11922584483346]
  Underwater Scene-level Enhancement (USE)<n>Uses a reliable scene representation for enhancement and a consistent enhancement target without paired enhanced 3D data.<n> Experiments on real underwater scenes show improved visibility and cross-view consistency while preserving reconstruction quality.
  arXiv  Detail & Related papers  (2026-08-28T07:34:13Z)
• Rethinking Conditional Generation for Underwater Salient Object Detection [69.38806804258452]
  A Conditional Generation Network (DCGNet) is designed to construct reliable conditional features for underwater saliency generation.<n>DCGNet significantly outperforms existing state-of-the-art methods, verifying its potential for complex underwater visual applications.
  arXiv  Detail & Related papers  (2026-07-02T07:48:06Z)
• Ocean4D: Generative Underwater 4D Reconstruction via Medium-Aware Video Diffusion [53.39217286144327]
  We propose a generative framework for underwater 4D reconstruction, named Ocean4D.<n>4D-GCC constructs 4D geometrically consistent conditioning with improved cross-frame coverage.<n>The Medium-Aware Block performs implicit medium-aware denoising in the latent diffusion process to stabilize underwater appearance.
  arXiv  Detail & Related papers  (2026-06-22T13:10:26Z)
• OceanSplat: Object-aware Gaussian Splatting with Trinocular View Consistency for Underwater Scene Reconstruction [4.325717217536016]
  OceanSplat is a novel 3D Gaussian Splatting-based approach for representing 3D geometry in underwater scenes.<n>We show that OceanSplat substantially outperforms existing methods for both scene reconstruction and restoration in scattering media.
  arXiv  Detail & Related papers  (2026-01-08T14:38:39Z)
• DGS-LRM: Real-Time Deformable 3D Gaussian Reconstruction From Monocular Videos [52.46386528202226]
  We introduce the Deformable Gaussian Splats Large Reconstruction Model (DGS-LRM)<n>It is the first feed-forward method predicting deformable 3D Gaussian splats from a monocular posed video of any dynamic scene.<n>It achieves performance on par with state-of-the-art monocular video 3D tracking methods.
  arXiv  Detail & Related papers  (2025-06-11T17:59:58Z)
• Plenodium: UnderWater 3D Scene Reconstruction with Plenoptic Medium Representation [31.47797579690604]
  We present Plenodium, a 3D representation framework capable of jointly modeling both objects and participating media.<n>In contrast to existing medium representations that rely solely on view-dependent modeling, our novel plenoptic medium representation incorporates both directional and positional information.<n>Experiments on real-world underwater datasets demonstrate that our method achieves significant improvements in 3D reconstruction.
  arXiv  Detail & Related papers  (2025-05-27T14:37:58Z)
• 3D-UIR: 3D Gaussian for Underwater 3D Scene Reconstruction via Physics Based Appearance-Medium Decoupling [30.985414238960466]
  3D Gaussian Splatting (3DGS) offers real-time rendering capabilities, but struggles with underwater inhomogeneous environments.<n>We propose a physics-based framework that disentangles object appearance from water medium effects.<n>Our approach achieves both high-quality novel view synthesis and physically accurate scene restoration.
  arXiv  Detail & Related papers  (2025-05-27T14:19:30Z)
• RUSplatting: Robust 3D Gaussian Splatting for Sparse-View Underwater Scene Reconstruction [9.070464075411472]
  This paper presents an enhanced Gaussian Splatting-based framework that improves both the visual quality and accuracy of deep underwater rendering.<n>We propose decoupled learning for RGB channels, guided by the physics of underwater attenuation, to enable more accurate colour restoration.<n>We release a newly collected dataset, Submerged3D, captured specifically in deep-sea environments.
  arXiv  Detail & Related papers  (2025-05-21T16:42:15Z)
• UW-GS: Distractor-Aware 3D Gaussian Splatting for Enhanced Underwater Scene Reconstruction [15.624536266709633]
  3D Gaussian splatting (3DGS) offers the capability to achieve real-time high quality 3D scene rendering.<n>However, 3DGS assumes that the scene is in a clear medium environment and struggles to generate satisfactory representations in underwater scenes.<n>We introduce a novel Gaussian Splatting-based method, UW-GS, designed specifically for underwater applications.
  arXiv  Detail & Related papers  (2024-10-02T13:08:56Z)
• EmerNeRF: Emergent Spatial-Temporal Scene Decomposition via Self-Supervision [85.17951804790515]
  EmerNeRF is a simple yet powerful approach for learning spatial-temporal representations of dynamic driving scenes. It simultaneously captures scene geometry, appearance, motion, and semantics via self-bootstrapping. Our method achieves state-of-the-art performance in sensor simulation.
  arXiv  Detail & Related papers  (2023-11-03T17:59:55Z)
• Physics Informed Neural Fields for Smoke Reconstruction with Sparse Data [73.8970871148949]
  High-fidelity reconstruction of fluids from sparse multiview RGB videos remains a formidable challenge. Existing solutions either assume knowledge of obstacles and lighting, or only focus on simple fluid scenes without obstacles or complex lighting. We present the first method to reconstruct dynamic fluid by leveraging the governing physics (ie, Navier -Stokes equations) in an end-to-end optimization.
  arXiv  Detail & Related papers  (2022-06-14T03:38:08Z)
• Learning Dynamic View Synthesis With Few RGBD Cameras [60.36357774688289]
  We propose to utilize RGBD cameras to synthesize free-viewpoint videos of dynamic indoor scenes. We generate point clouds from RGBD frames and then render them into free-viewpoint videos via a neural feature. We introduce a simple Regional Depth-Inpainting module that adaptively inpaints missing depth values to render complete novel views.
  arXiv  Detail & Related papers  (2022-04-22T03:17:35Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.