MonoStereoFusion

This repository contains the code and dataset for our paper Yuhua Xu, Xiaoli Yang, Yushan Yu, Wei Jia, Zhaobi Chu, Yulan Guo. Depth Estimation by Combining Binocular Stereo and Monocular Structured-Light, CVPR 2022.

Requirements

The code has been tested with PyTorch 1.8.1 and Cuda 10.1.

Dataset

We provided the MonoBinoStereo dataset, check /MonoBinoStereo-dataset

The dataset includes:

  • passive stereo images
  • stereo images with asymmetric textures (the left images are passive, and the right images are with speckles)
  • depth maps generated from the structured light subsystem. These depth maps are aligned with the left images.
  • the ground truth disparity maps generated by the space-time stereo methods
  • parameters of cameras

Pre-trained model

We provide pre-trained models RAFT-OM-G and RAFT-O, under the folder /trained-models

Evaluation

We provided a script to get the MonoBinoStereo dataset benchmark result.

evaluation.py provides an example usage.

Download

OneDrive Link: https://1drv.ms/u/s!AhORN5PjOtgJgTFmL8LsM94BSLvu?e=faAQjQ

License agreement

This dataset is made freely available to academic and non-academic entities for non-commercial purposes such as academic research, teaching, scientific publications, or personal experimentation.

Acknowledgements

Part of the code is adopted from the previous works: RAFT, PSMNet. We thank the original authors for their contributions.