This repository contains all the code to reproduce the experiments in our paper Siamese CNNs for RGB-LWIR Disparity Estimation. It is separated into three modules.
- Patch Generator: Generates the disparity locations (center of patches) for training, validation and testing set.
- Rectification: Rectifies images of the St-Charles dataset.
- Stereo: Generates dataset based on the chosen fold and contains scripts to train and test the model.
- Shared: Contains configuration file which all modules use.
@InProceedings{Beaupre_2019_CVPR_Workshops,
author = {Beaupre, David-Alexandre and Bilodeau, Guillaume-Alexandre},
title = {Siamese CNNs for RGB-LWIR Disparity Estimation},
booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
month = {June},
year = {2019}
}
Please refer to README files in each module for more details. Make sure to change paths and other variables in the config file to your own values.
- Rectify the images from the St-Charles dataset with the Rectification module.
- Generate the dataset from the dataset.py script in the stereo module.
- Generate patch locations with the Patch Generator module.
- Train or test the model with the scripts in the Stereo module.
Simply put both datasets in a folder named "litiv" where all your datasets are located.
For any comments, questions or concerns, feel free to contact me at [email protected]
See the LICENSE file for more details.