license: bsd-3-clause
task_categories:
- depth-estimation
The official implementation is available on GitHub.
Zero-Shot Depth from Defocus
Yiming Zuo* · Hongyu Wen* · Venkat Subramanian* · Patrick Chen · Karhan Kayan · Mario Bijelic · Felix Heide · Jia Deng
(*Equal Contribution)
Princeton Vision & Learning Lab (PVL)
Paper · Project Page · Code
Overview
Depth from Defocus (DfD) is the task of estimating a dense metric depth map from a focus stack. Unlike previous works overfitting to a certain dataset, this paper focuses on the challenging and practical setting of zero-shot generalization.
We first propose a new real-world DfD benchmark ZEDD (released under CC BY 4.0), which contains 8.3x more scenes and significantly higher quality images and ground-truth depth maps compared to previous benchmarks. We also build the Infinigen Defocus synthetic dataset on top of Infinigen Indoors. Infinigen is a procedural system for generating photorealistic indoor scenes.
Infinigen uses Blender for scene composition and rendering. Blender provides native support for camera aperture and focus distance, and supports synthesizing defocus effects during ray tracing using a thin-lens camera model. This makes Blender suitable for generating realistic focus stacks with physically accurate defocus blur.
We modify the Infinigen generation pipeline so that, for each scene, it renders multiple images from the same camera pose while varying the aperture size and focus distance. Specifically, we render images using 5 aperture settings (F1.4/2.0/2.8/4.0/5.6), 9 focus distances (0.8/1.2/1.7/2.3/3.0/3.8/4.7/6.0/8.0m), and one additional all-in-focus image, resulting in 5 × 9 + 1 = 46 images per scene.
We use the rendered depth map of the all-in-focus image as the ground-truth depth. In total, we generate 500 scenes and manually reject the scenes with degenerated object layout or suboptimal camera placement, resulting in 200 scenes with the highest visual quality.
Citation
@article{ZeroShotDepthFromDefocus,
author = {Zuo, Yiming and Wen, Hongyu and Subramanian, Venkat and Chen, Patrick and Kayan, Karhan and Bijelic, Mario and Heide, Felix and Deng, Jia},
title = {Zero-Shot Depth from Defocus},
journal = {arXiv preprint arXiv:2603.26658},
year = {2026},
url = {https://arxiv.org/abs/2603.26658}
}