Official repository for BMVC 2022 paper: Global Proxy-based Hard Mining for Visual Place Recognition
This paper introduces a novel approach to deep representation learning for Visual Place Recognition. Our method utilizes global hard mini-batch sampling based on proxies, achieved by adding an end-to-end trainable branch to the network that generates efficient place descriptors. These descriptors are then used to construct a global index that captures the similarities between all places in the dataset, enabling highly informative mini-batch sampling at each training iteration. The proposed technique can be used with existing pairwise and triplet loss functions at minimal additional memory and computation cost.
GPM is effective for a wide range of mini-batch sizes, with more impact when smaller mini-batches are used for training. This is of great importance when training hardware resources are limited.
Code coming soon
Use the following bibtex code to cite our paper
@inproceedings{Ali-Bey_2022_BMVC,
author = {Amar Ali-Bey and Brahim Chaib-draa and Philippe Giguere},
title = {Global Proxy-based Hard Mining for Visual Place Recognition},
booktitle = {33rd British Machine Vision Conference 2022, {BMVC} 2022, London, UK, November 21-24, 2022},
publisher = {{BMVA} Press},
year = {2022},
url = {https://bmvc2022.mpi-inf.mpg.de/0958.pdf}
}