Live User-Guided Depth Map Estimation for Single Images

by Márcio C. F. Macedo and Antônio L. Apolinário Jr.

Introduction

This is a C++ application for user-guided, real-time depth map estimation in single images. Technical details are provided in our paper accepted in the Journal on Real-Time Image Processing.

The provided source code was tested using the following libraries:

• Eigen 3.3.9;
• OpenCV 4.5.1;
• CUDA 11.2;

The application receives as input the following console arguments:

• -i image.extension: to provide an input image for the application;
• -a image.extension: to provide an initial annotation for the input image;
• --live: to enable live depth annotation;

Once the application is started, one can interact with the depth map estimation using the following keys:

• Press '0' to annotate the "Edited Image" with depth 0;
• Press '1' to annotate the "Edited Image" with depth 64;
• Press '2' to annotate the "Edited Image" with depth 128;
• Press '3' to annotate the "Edited Image" with depth 192;
• Press '4' to annotate the "Edited Image" with depth 255;
• Press '+' to increase the scribble radius;
• Press '-' to decrease the scribble radius;
• Press 't' or 'T' to print the processing time demanded by one frame into the console;
• Press 's' or 'S' to save the estimated depth map, the annotated image and the artistic depth-based effect;
• Press 'd' or 'D' to start the depth map estimation process (this process is automatic if --live is enabled);
• Press 'g' or 'G' to visualize the desaturation effect (the output image is automatically updated if --live is enabled);
• Press 'h' or 'H' to visualize the haze effect (the output image is automatically updated if --live is enabled);
• Press 'b' or 'B' to visualize the refocus effect (the output image is automatically updated if --live is enabled);

Citation

The provided source codes are in public domain and can be downloaded for free. If this work is useful for your research, please consider citing:

@article{Macedo2021,
author={Macedo, M{\'a}rcio C. F. and Apolin{\'a}rio, Ant{\^o}nio L.},
title={Live User-Guided Depth Map Estimation for Single Images},
journal={Journal of Real-Time Image Processing},
year={2021},
month={Jan},
day={13},
issn={1861-8219},
doi={10.1007/s11554-020-01055-x},
}