Infinite Leagues Under the Sea: Photorealistic 3D Underwater Terrain Generation by Latent Fractal Diffusion Models
Tianyi Zhang Weiming Zhi Joshua Mangelson Matthew Johnson-Roberson · 2025
Read Infinite Leagues Under the Sea: Photorealistic 3D Underwater Terrain Generation by Latent Fractal Diffusion Models by Tianyi Zhang Weiming Zhi Joshua Mangelson Matthew Johnson-Roberson (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-06784.txt.
Description
Tianyi Zhang and three colleagues present a 2025 paper on generating realistic underwater terrain for film, games, and robot training. Generic image generators fail here because seafloor photos are rare in their training data, so the results look fake. The team's model, DreamSea, learns instead from real survey images gathered by underwater robots.
From those noisy surveys they extract shape and meaning using vision foundation models. Then they train a diffusion generator that paints seafloor views with depth, steered by fractal-based codes that keep large areas coherent. The views are fused into a three-dimensional scene that renders convincingly from new angles. Tests showed large, varied, photorealistic seabeds, a resource for anyone who needs an ocean floor without sailing to one.
Catalog record
| CREATOR | Tianyi Zhang Weiming Zhi Joshua Mangelson Matthew Johnson-Roberson |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 31 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-06784