h-tr / atten_patch

Atten Patch: Visual Place Recognition using Local Descriptors

Overview

Atten Patch is a research project aimed at enhancing visual place recognition through the utilization of local descriptors. The project focuses on leveraging key features to optimize the algorithm's speed without compromising precision. Additionally, it incorporates training-based methods to further augment accuracy.

Getting Started

To get started with the Atten Patch project, follow these steps:

1. Installation: Clone the repository to your local machine.

git clone https://github.com/H-tr/atten_patch.git
cd atten_patch

2. Dependencies: Install the required dependencies using pip and the provided requirements.txt file.

pip install -r requirements.txt

You can install the pretrained models and checkpoints into pretrained_models

mim download mmsegmentation --config segformer_mit-b4_8xb2-160k_ade20k-512x512 --dest pretrained_models

This will install all necessary dependencies for the project.

1. Configuration: You can configurate the project in config.

Usage

Evaluate on SPED dataset:

# Command to run the project
$ python main.py --dataset SPED --config config/vpr_bench_config.yaml

License

This project is licensed under the MIT License - see the LICENSE file for details.

The MIT License is a permissive open-source license that allows for the modification and distribution of the software while requiring a copy of the original license and disclaimer of warranty to be included in derivative works. You can find the full text of the license in the LICENSE file within this repository.

Acknowledgments

We would like to extend our gratitude to the following individuals and organizations for their contributions, guidance, and support:

• [Zhang Dongshuo]: Kindly provide source code