ZeroStereo: Zero-shot Stereo Matching from Single Images

TL;DR

ZeroStereo introduces a novel pipeline that synthesizes stereo images from single images, enabling zero-shot stereo matching with improved generalization to real-world data without additional training.

Contribution

The paper presents a new zero-shot stereo matching pipeline using image synthesis, diffusion inpainting, and confidence estimation, advancing zero-shot generalization in stereo vision.

Findings

Achieves state-of-the-art zero-shot generalization on multiple datasets.

Requires only a small dataset comparable to Scene Flow.

Outperforms previous methods in real-world scenarios.

Abstract

State-of-the-art supervised stereo matching methods have achieved remarkable performance on various benchmarks. However, their generalization to real-world scenarios remains challenging due to the scarcity of annotated real-world stereo data. In this paper, we propose ZeroStereo, a novel stereo image generation pipeline for zero-shot stereo matching. Our approach synthesizes high-quality right images from arbitrary single images by leveraging pseudo disparities generated by a monocular depth estimation model. Unlike previous methods that address occluded regions by filling missing areas with neighboring pixels or random backgrounds, we fine-tune a diffusion inpainting model to recover missing details while preserving semantic structure. Additionally, we propose Training-Free Confidence Generation, which mitigates the impact of unreliable pseudo labels without additional training, and Adaptive Disparity Selection, which ensures a diverse and realistic disparity distribution while preventing excessive occlusion and foreground distortion. Experiments demonstrate that models trained with our pipeline achieve state-of-the-art zero-shot generalization across multiple datasets with only a dataset volume comparable to Scene Flow. Code: https://github.com/Windsrain/ZeroStereo.