BenthicFlow: Generating Extensible Underwater Environments via Flow Matching
2026-08-24 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors created BenthicFlow, a new computer program that generates realistic 3D images of underwater scenes by combining textures and depth maps all at once. Unlike older methods that piece together smaller parts afterward, their approach makes large, continuous underwater landscapes that look natural and accurate. Their system uses a special sampling technique to smoothly blend overlapping image sections, and then turns these images into detailed 3D surfaces. Tests show BenthicFlow can produce big, coherent underwater scenes that match real-world data from different locations.
3D scene understandingunderwater computer visiongenerative modelsflow-matching modeldepth mapsRGBD mosaicsMultiDiffusion samplingGaussian surfelsdomain gap
Authors
Joaquín Figueira, Camile Lendering, Manfred Gonzalez-Hernandez, Giacomo D'Amicantonio, Erkut Akdag, Egor Bondarev
Abstract
Computer vision applications for 3D scene understanding in underwater environments remain challenging due to the lack of high-quality 3D data and the inability of surface-trained models to generalize to underwater scenes. To address this challenge, an emerging trend is to employ generative models to close the data domain gap. However, existing methods assemble large scenes by stitching independently generated tiles post hoc with separately trained models, while demonstrating heterogeneous landscapes only within individual survey sites. We introduce BenthicFlow, a unified framework based on a single conditional flow-matching model that jointly generates aligned textures and depth maps. A MultiDiffusion-inspired sampling procedure reconciles overlapping windows throughout the generative trajectory, enabling spatially extensible RGBD mosaics without a separate stitching model. The generated mosaics are subsequently lifted into explicit 3D benthic environments using surface-aligned Gaussian surfels. Experiments across geographically distinct survey sites demonstrate that BenthicFlow preserves site-specific appearance while generating coherent, large-scale 3D scenes that closely match the target distributions. Code and trained models are available at https://github.com/jacomof/BenthicFlow.