Visual Place Recognition: A Tutorial

Work in progress: This repository provides the example code from our paper "Visual Place Recognition: A Tutorial". The code performs VPR on the GardensPoint day_right--night_right dataset. Output is a plotted pr-curve, matching decisions, two examples for a true-positive and a false-positive matching, and the AUC performance, as shown below.

If you use our work for your academic research, please refer to the following paper:

@article{SchubertVisual,
    title={Visual Place Recognition: A Tutorial},
    author={Stefan Schubert and Peer Neubert and Sourav Garg and Michael Milford and Tobias Fischer},
    journal={arXiv 2303.03281},
    year={2023},
}

How to run the code

Online

We provide the Jupyter Notebook demo.ipynb that runs the code, shows intermediate results, and provides some additional explanations.

Local

python3 demo.py

The GardensPoints Walking dataset will be downloaded automatically. You should get an output similar to this:

python3 demo.py
===== Load dataset
===== Load dataset GardensPoint day_right--night_right
===== Compute local DELF descriptors
===== Compute holistic HDC-DELF descriptors
===== Compute cosine similarities S
===== Match images
===== Evaluation

===== AUC (area under curve): 0.7419284432277942
===== R@100P (maximum recall at 100% precision): 0.36
===== recall@K (R@K) -- R@1: 0.85 , R@5: 0.925 , R@10: 0.945

Precision-recall curve	Matchings M	Examples for a true positive and a false positive

Requirements

The code was tested with the following library versions. Note that Tensorflow or PyTorch is only required if the corresponding image descriptor is used.

pipreqs VPR_Tutorial/ --print

• matplotlib==3.1.2
• numpy==1.17.4
• Pillow==9.5.0
• scikit_image==0.19.3
• scipy==1.3.3
• skimage==0.0
• tensorflow==2.12.0
• tensorflow_hub==0.13.0
• torch==1.13.1+cpu
• torchvision==0.14.1+cpu

List of existing open-source implementations for VPR (work in progress)

Descriptors

Holistic descriptors

AlexNet	code*	paper
AMOSNet	code	paper
DELG	code	paper
DenseVLAD	code	paper
HDC-DELF	code	paper
HybridNet	code	paper
NetVLAD	code	paper

Local descriptors

D2-Net	code	paper
DELF	code	paper
LIFT	code	paper
Patch-NetVLAD	code	paper
R2D2	code	paper
SuperPoint	code	paper

Local descriptor aggregation

DBoW2	code	paper
HDC (Hyperdimensional Computing)	code	paper
iBoW (Incremental Bag-of-Words) / OBIndex2	code	paper
VLAD (Vector of Locally Aggregated Descriptors)	code*	paper

Sequence methods

Delta Descriptors	code	paper
MCN	code	paper
OpenSeqSLAM	code*	paper
OpenSeqSLAM 2.0	code	paper
OPR	code	paper
SeqConv	code	paper
SeqNet	code	paper
SeqVLAD	code	paper
VPR	code	paper

Misc

ICM	code	paper	Graph optimization of the similarity matrix S
SuperGlue	code	paper	Local descriptor matching

*Third party code. Not provided by the authors. Code implements the author's idea or can be used to implement the authors idea.