Single-Image Depth Perception in the Wild

TL;DR

This paper introduces a new dataset and a simpler, more effective algorithm for estimating metric depth from single images in unconstrained environments, advancing the field of in-the-wild depth perception.

Contribution

It presents a novel dataset with relative depth annotations and a new algorithm that outperforms existing methods in single-image depth estimation.

Findings

Our algorithm outperforms state-of-the-art methods.

Combining RGB-D data with relative depth annotations improves accuracy.

The dataset enables better training and evaluation of depth perception models.

Abstract

This paper studies single-image depth perception in the wild, i.e., recovering depth from a single image taken in unconstrained settings. We introduce a new dataset "Depth in the Wild" consisting of images in the wild annotated with relative depth between pairs of random points. We also propose a new algorithm that learns to estimate metric depth using annotations of relative depth. Compared to the state of the art, our algorithm is simpler and performs better. Experiments show that our algorithm, combined with existing RGB-D data and our new relative depth annotations, significantly improves single-image depth perception in the wild.