Out-of-focus: Learning Depth from Image Bokeh for Robotic Perception

TL;DR

This paper introduces a novel method for depth estimation in robotics by using sequences of focal images captured at different focus settings, leveraging natural blur cues to improve accuracy over traditional single-image methods.

Contribution

The work presents a new approach that uses multiple focal images and a CNN to enhance depth estimation, supported by a custom dataset and comparative evaluations.

Findings

Stacked focal images outperform single RGB images in depth estimation.

The proposed method improves accuracy on both standard and custom datasets.

Using focus variation provides richer depth cues than traditional RGB methods.

Abstract

In this project, we propose a novel approach for estimating depth from RGB images. Traditionally, most work uses a single RGB image to estimate depth, which is inherently difficult and generally results in poor performance, even with thousands of data examples. In this work, we alternatively use multiple RGB images that were captured while changing the focus of the camera's lens. This method leverages the natural depth information correlated to the different patterns of clarity/blur in the sequence of focal images, which helps distinguish objects at different depths. Since no such data set exists for learning this mapping, we collect our own data set using customized hardware. We then use a convolutional neural network for learning the depth from the stacked focal images. Comparative studies were conducted on both a standard RGBD data set and our own data set (learning from both single and multiple images), and results verified that stacked focal images yield better depth estimation than using just single RGB image.