bathy_nn_learning

Introduction

This repository is intended for loop closure detection and feature matching in the context of Multibeam Echo Sounders (MBES).

Dependencies

• Python 3.8
• PyTorch 1.10
• PyTorch Geometric 2.0
• opencv, wandb, shapely, open3d, AUVLib (original repo)

Install the June 2022 version of AUVLib here:

git clone -b extended_bm git@github.com:ignaciotb/auvlib.git

For details, please refer to requirements.txt (for pip) or environment.yml (for conda).

Recommended for baseline: PCL 1.10 or its python binding

Data

Datasets are available for download here datasets

Usage

Step 1: Run scripts/parse_cereal.py to parse cereal data.

Step 2.1 - 2.3: Run other scripts in scripts/ to create datasets.

Step 3: Run train.py to train a model. (Modify param.py properly.)

Step 4: Run scripts in test/ to evaluate the model.

.
├── data               # datasets
│   ├── Circle100      # training set
│   │   ├── raw        # raw training set
│   │   └── processed  # processed training set
│   ├── Circle100Valid
│   │   └── ...
│   └── Circle100Test
│       └── ...
├── scripts     # scripts for data processing
├── utils       # utility functions 
├── test        # testing scripts
├── models.py   # model implementation
├── dataset.py  # dataset implementation
├── param.py    # parameters and configurations
└── train.py    # training script

Citation

If you find our work useful, please consider citing:

@inproceedings{tan2023data,
    title={Data-driven loop closure detection in bathymetric point clouds for underwater {SLAM}},
    author={Tan, Jiarui and Torroba, Ignacio and Xie, Yiping and Folkesson, John},
    booktitle={2023 IEEE International Conference on Robotics and Automation (ICRA)},
    pages={3131--3137},
    year={2023},
    organization={IEEE}

}

Acknowledgment

Part of the code is based on some examples in PyTorch Geometric.