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
Figures & tables
Figure 1. We present NemoSplat, a novel feed-forward model that directly reconstructs photorealistic underwater scenes from uncalibrated image sequences, robustly handling numerous dynamic objects and water-induced degradation. A teaser image showing the main result and concept of the paper.
Figure 2. Overview of our NemoSplat framework. It leverages the Visual Geometry Transformer to extract robust spatiotemporal features from uncalibrated underwater sequences. Subsequently, a shared feature encoder feeds these representations into specialized decoders to jointly estimate camera poses, depth, dynamic semantic masks (optionally refined via grounded text prompts), water backscatter and attenuation parameters, and intrinsic 4D Gaussians. Overview figure.
Figure 3. Dynamic-Semantic Mask Generation. The Promptable Dynamic Disentangler ( FM ) refines the predicted dynamic mask via logit adjustment using an optional text prompt-guided semantic map, effectively revising a highly accurate mask for downstream 4D modeling. A technical flowchart showing the Dynamic-Semantic Mask Generation pipeline.
Figure 4. Examples sequences of our proposed evaluation dataset. It includes synthetic scenes with precise ground-truth annotations and diverse in-the-wild underwater sequences for comprehensive evaluation. Sampled frames and depth maps from the evaluation dataset.
Method
BoulderShore
Coral
Deepsea
Average
Metrics
ATE ↓
PSNR ↑
SSIM ↑
LPIPS ↓
ATE ↓
PSNR ↑
SSIM ↑
LPIPS ↓
ATE ↓
PSNR ↑
SSIM ↑
LPIPS ↓
ATE ↓
PSNR ↑
SSIM ↑
LPIPS ↓
VGGT ( Wang et al., 2025a )
0.19
-
-
-
0.23
-
-
-
5.93
-
-
-
2.12
-
-
-
StreamVGGT ( Zhuo et al., 2025 )
0.40
-
-
-
0.79
-
-
-
5.01
-
-
-
2.06
-
-
-
WildGS-SLAM ( Zheng et al., 2025 )
0.12
18.80
0.60
0.45
4.97
17.01
0.55
0.52
0.81
29.86
0.81
0.35
1.96
21.89
0.65
0.44
Droid-W ( Li et al., 2026 )
0.11
20.04
0.71
0.44
4.65
18.56
0.58
0.56
0.78
31.93
0.83
0.30
1.85
23.51
0.71
0.43
YoNoSplat ( Ye et al., 2026 )
0.39
17.81
0.46
0.55
0.27
17.63
0.44
0.56
×
×
×
×
×
×
×
×
Table 1. Quantitative evaluation on synthetic sequences. Best results are bold , second best are underlined . “-” indicates that the method does not support rendering, and “ × ” denotes failure due to out of memory. ATE is reported in meters.
Method
Metrics
01
02
03
04
05
06
07
08
09
10
11
12
13
14
Avg.
WildGS-SLAM ( Zheng et al., 2025 )
PSNR ↑
19.31
18.13
20.85
19.92
19.24
21.75
19.79
16.08
21.40
19.38
20.79
21.64
16.29
13.50
19.15
SSIM ↑
0.56
0.65
0.74
0.62
0.52
0.74
0.62
0.42
0.68
0.49
0.69
0.78
0.58
0.37
0.60
LPIPS ↓
0.50
0.52
0.38
0.37
0.44
0.42
0.44
0.50
0.35
0.36
0.28
0.25
0.41
0.58
0.41
Droid-W ( Li et al., 2026 )
PSNR ↑
20.59
19.83
18.61
18.73
17.77
20.50
17.05
16.71
18.43
19.05
17.90
21.05
13.20
14.32
18.12
SSIM ↑
0.73
0.69
0.51
0.55
0.40
0.68
0.49
0.50
0.60
0.46
0.63
0.72
0.50
0.43
0.56
LPIPS ↓
0.37
0.39
0.46
0.37
0.52
0.51
0.58
0.42
0.43
0.39
0.36
0.27
0.50
0.57
0.44
Table 2. Quantitative evaluation of novel view synthesis on 14 real-world aquatic sequences. We report PSNR, SSIM, and LPIPS metrics across all scenes alongside their overall averages. The best results are bold , and the second best are underlined .
Figure 5. Restored images on SeaThru-NeRF ( Levy et al., 2023 ) dataset. Our approach recovers more natural colors while preserving clearer details in distant regions. Figures showing restored images on SeaThru-NeRF dataset.
Figure 6. Qualitative comparison of novel view synthesis on highly degraded real-world aquatic sequences. While conventional feed-forward modules and SLAM-based baselines suffer from severe ghosting, topological blurring, and visual artifacts caused by unconstrained dynamic entities (e.g., schools of swimming fish), NemoSplat explicitly decouples transient motion from the static topology. This disentanglement enables the synthesis of crisp, temporally consistent, and artifact-free novel views. Figures showing rendering results of baseline models and NemoSplat.
Figure 7. Reconstruction results on real-world aquatic sequences. Our method disentangles static Gaussian models from dynamic underwater scenes, providing comprehensive outputs including RGB rendering images, depth maps and dynamic masks.
Underwater 4D reconstruction remains challenging due to the coupling between degraded light transport in participating media and dynamic water variations. Most existing Methods are developed under in-air assumptions and do not explicitly account for underwater absorption and backscatter. Additionally, near-static assumptions make these approaches sensitive to drifting particles and dynamic distractors , leading to unstable geometry and inconsistent cross-view results. To address these issues, we propose a generative framework for underwater 4D reconstruction, named Ocean4D, which is built on two complementary components. Specifically, 4D-GCC constructs 4D geometrically consistent conditioning with improved cross-frame coverage, while the Medium-Aware Block performs implicit medium-aware denoising in the latent diffusion process to stabilize underwater appearance under absorption and scattering. Given a monocular video and target cameras, our method generates videos along the target trajectories while preserving global structure and cross-view consistency. Extensive experiments on both dynamic and static underwater benchmarks demonstrate state-of-the-art performance on underwater reconstruction.
Yuqiang Huang, Yuxi Wang, Junyu Dong +1
Faculty of Information Science and Engineering, Ocean University of China · Institute of Automation, Chinese Academy of Sciences
Recent monocular 3D Gaussian Splatting (3DGS) streaming reconstruction methods have achieved impressive performance by balancing reconstruction quality and efficiency. However, extending these frameworks to underwater scenes remains challenging due to severe visual degradation, such as light attenuation and scattering, which degrades camera pose tracking and distorts scene geometry. To address these challenges, we propose AquaFlow, a monocular Gaussian Splatting streaming reconstruction framework for efficient and high-fidelity underwater reconstruction. Specifically, AquaFlow fine-tunes a 3D vision foundation model on large-scale underwater data for robust pose and pointmap estimation, and introduces a medium-guided incremental Gaussian initialization strategy for streaming mapping. Furthermore, we develop a streaming-compatible hybrid scene representation that integrates structured, distance-conditioned neural Gaussians with a physics-inspired optical model to compensate for underwater image formation effects, enabling accurate scene reconstruction. We evaluate AquaFlow on a comprehensive dataset of 62 diverse underwater trajectories, collected from both public benchmarks and in-the-wild web videos across various scales. Extensive experiments demonstrate that AquaFlow achieves state-of-the-art tracking and rendering performance, reducing average localization error by 13.2% and improving PSNR by 4.74 dB compared to WaterSplat-SLAM.
Yingxiang Xu, Kerui Ren, Wenqi Guo +4
Zhejiang University · Shanghai AI Laboratory · Shanghai Jiao Tong University +3
Underwater scene reconstruction is essential for immersive exploration of aquatic environments, yet remains challenging due to complex participating-media effects such as absorption and scattering, as well as the limited field of view (FoV) of conventional cameras. Although combining panoramic imaging with 3D Gaussian Splatting (3DGS) offers a promising direction for photorealistic underwater rendering, traditional 3DGS struggles with both spherical projection distortion and underwater medium degradation. In this paper, we propose \textbf{Underwater360}, a physics-informed omnidirectional 3DGS framework for underwater panoramic scene reconstruction. First, we introduce an Omnidirectional Gaussian Splatting module that performs ray casting directly in spherical camera space instead of relying on 2D projection approximations, thereby reducing geometric distortions under 360∘ FoV. Second, we design a physics-based appearance-medium modeling architecture with pose-conditioned appearance embeddings to explicitly decouple intrinsic scene radiance from depth-dependent backscatter and attenuation, enabling physically grounded scene appearance restoration. Finally, we establish a new panoramic underwater benchmark dataset containing both synthetic and real-world scenes. Extensive experiments demonstrate that Underwater360 achieves superior performance in underwater novel view synthesis and scene appearance restoration, delivering improved rendering quality and cross-view consistency in complex underwater environments. The code and datasets are released at https://github.com/SwcK423/Underwater360
Jiangbei Hu, Weichao Song, Shibo Yu +8
School of Software, Dalian University of Technology · College of Computing and Data Science, Nanyang Technological University