AI-Place-Recognition

For all main contributors, please check contributing.

Introduction

This repository is dedicated to research and development of AI-based place recognition.

How To Use

Clone

Clone this GitHub repository:

git clone https://github.com/SKHU-Adventure/ai-place-recognition.git
cd ai-place-recognition

Requirements

The main branch works with CUDA 12.1, CUDNN 8.9.2, NCCL 2.18.3 with Python 3.8. Refer to a document for pre-setting and then install requirements:

pip install -r requirements.txt

Prepare Datasets

1. Prepare dataset for training:

For public datasets, you may refer to following links.

• Nordland
• Tokyo (available on request)

How to Train

1. Create a directory for your experiment (e.g., experiments/sample)

2. Write a setup.ini file in the directory (e.g., experiments/sample/setup.ini).

3. Run:

python3 train.py [EXPERIMENT_DIR]

where [EXPERIMENT_DIR] is the directory created above.

How to Evaluate

How to Demo

Contributing

Main contributors:

• Mujae Park, mujae9837[at]gmail.com
• Younguk Jeon, jayiuk987[at]gmail.com
• Heeju Cha, chaheeju01[at]gmail.com
• Younah Kim, kkiwiio[at]gmail.com

Advisior:

• Sangyun Lee, sylee[at]skhu.ac.kr