FoundationGeo: Learning Spatial Pixel-Wise Fields for Monocular Metric Geometry

2026-07-13

AI Summary

This work addresses the sharp degradation in metric geometric reconstruction performance when monocular images are captured under camera intrinsics—such as focal length—that differ between training and testing distributions. To tackle this issue, the authors propose a two-stage approach: first, they learn an affine-invariant, high-fidelity geometric representation using DINOv3 pretrained on tens of millions of multi-domain images; second, they introduce a lightweight, pixel-wise calibration field that spatially adapts scale and ray direction for metric-consistent 3D point map reconstruction. The study systematically identifies insufficient coverage of camera intrinsics as a critical bottleneck in zero-shot metric generalization and mitigates this gap by synthesizing diverse focal lengths in Blender. Evaluated across seven benchmarks, the method outperforms heavier baselines by over 5.2% on average, substantially improving cross-domain zero-shot generalization.

Abstract

We present FoundationGeo, a two-stage framework that explicitly bridges relative and metric prediction via spatial calibration and principled data design. Stage 1 learns a high-fidelity, affine-invariant geometry model by initializing with DINOv3 and training on a curated 10.2M-sample multi-domain corpus with complementary local-detail supervision, yielding sharp boundaries and strong cross-domain generalization. Stage 2 moves beyond global scaling by introducing lightweight pixel-wise calibration fields for metric estimation: a scale field for spatially varying metric alignment and a ray-direction correction field that mitigates directional bias in point-map geometry, together producing metrically consistent 3D point maps. Beyond model design, we identify camera intrinsic coverage, especially focal length distribution mismatch between training and test data, as a key bottleneck for zero-shot metric generalization: performance drops sharply when test intrinsics fall outside the training distribution. To address this, we synthesize additional training data across diverse focal lengths using a Blender-based data engine, repairing under-covered focal regimes and improving robustness under intrinsic shift. Extensive zero-shot evaluations across seven benchmarks show that FoundationGeo significantly strengthens cross-domain robustness, staying near the top across diverse domains while avoiding the sharp cross-domain performance drops observed in other methods. This consistency translates into the best overall performance, surpassing heavier baselines by over 5.2% on average.

Problem

• monocular metric geometry
• zero-shot generalization
• camera intrinsics
• cross-domain robustness
• focal length distribution

Innovation

• pixel-wise calibration fields
• metric geometry
• zero-shot generalization
• camera intrinsic robustness
• multi-domain training