G$^2$VLM: Geometry Grounded Vision Language Model with Unified 3D Reconstruction and Spatial Reasoning

TL;DR

G$^2$VLM is a geometry grounded vision-language model that integrates 3D reconstruction and spatial reasoning, improving robustness in spatial understanding tasks by leveraging learned 3D visual geometry features.

Contribution

The paper introduces G$^2$VLM, a unified model that combines 3D reconstruction and spatial reasoning in vision-language tasks, trained on multi-view data without requiring extensive annotations.

Findings

Achieves state-of-the-art or competitive results in 3D reconstruction and spatial reasoning tasks.

Effectively leverages multi-view image and video data for training.

Demonstrates the model's potential for future applications like 3D scene editing.

Abstract

Vision-Language Models (VLMs) still lack robustness in spatial intelligence, demonstrating poor performance on spatial understanding and reasoning tasks. We attribute this gap to the absence of a visual geometry learning process capable of reconstructing 3D space from 2D images. We present G$^2$VLM, a geometry grounded vision-language model that bridges two fundamental aspects of spatial intelligence: spatial 3D reconstruction and spatial understanding. G$^2$VLM natively leverages learned 3D visual geometry features to directly predict 3D attributes and enhance spatial reasoning tasks via in-context learning and interleaved reasoning. Our unified design is highly scalable for spatial understanding: it trains on abundant multi-view image and video data, while simultaneously leveraging the benefits of 3D visual priors that are typically only derived from hard-to-collect annotations. Experimental results demonstrate G$^2$VLM is proficient in both tasks, achieving comparable results to state-of-the-art feed-forward 3D reconstruction models and achieving better or competitive results across spatial understanding and reasoning tasks. By unifying a semantically strong VLM with low-level 3D vision tasks, we hope G$^2$VLM can serve as a strong baseline for the community and unlock more future applications, such as 3D scene editing.