Wat3R: Underwater 3D Geometry Learning Without Annotations

Forum topic · 小凯 · 2026-07-11

Summary

Wat3R is a cross-domain semi-supervised learning framework that adapts feed-forward 3D reconstruction models from aerial to underwater scenes without requiring any annotated underwater data. Underwater 3D geometry estimation is challenging due to light attenuation, scattering, and the lack of large-scale high-quality 3D annotations, which makes existing densely supervised methods impractical. Wat3R uses a teacher-student architecture trained on abundant unlabeled real underwater video footage, and introduces a cross-view consistency loss that leverages geometric cues from other viewpoints to compensate for information degradation caused by underwater attenuation and scattering. The authors also present Water3D, a new dataset covering diverse water bodies and underwater scenes for benchmarking geometry tasks. Experiments show Wat3R outperforms state-of-the-art methods on underwater multi-view depth estimation and point cloud reconstruction. Paper: arXiv 2507.08184.

Overview

Field: Computer Vision
Authors: Jiangwei Ren, Xingyu Jiang, Zijie Song
arXiv: 2507.08184

Key Points

Estimating 3D geometry underwater is uniquely difficult because of light attenuation, scattering, and the absence of large-scale, high-quality 3D annotations.
Existing methods rely on massive dense annotations, which are impractical to collect in underwater settings.
Wat3R is a cross-domain semi-supervised learning framework that adapts feed-forward 3D reconstruction models trained in air to underwater scenes.
It requires no annotated underwater data: a teacher-student architecture learns robust geometry representations purely from abundant unlabeled real underwater video footage.
A cross-view consistency loss leverages geometric cues from other viewpoints to compensate for information degradation caused by underwater attenuation and scattering.
The authors introduce Water3D, a new dataset spanning diverse water bodies and underwater scenes, designed for benchmarking underwater geometry tasks.

Results

Experiments demonstrate that Wat3R outperforms current state-of-the-art methods on both underwater multi-view depth estimation and point cloud reconstruction.