This repository contains the code for our NeurIPS 2020 publication Soft Contrastive Learning for Visual Localization.
The corresponding models, training/testing image lists and a movie with visual results can be downloaded here.
This code was tested using TensorFlow 1.10.0 and Python 3.5.6.
It uses the following git repositories as dependencies:
The training data can be downloaded using:
Models used in the paper
| Name | Model |
|---|---|
| Off-the-shelf | offtheshelf |
| Triplet trained on Pittsburgh | pittsnetvlad |
| Triplet | triplet_xy_000 |
| Quadruplet | quadruplet_xy_000 |
| Lazy triplet | ha0_lolazy_triplet_muTrue_renone_vl64_pca_eccv_002 |
| Lazy quadruplet | ha0_lolazy_quadruplet_muTrue_renone_vl64_pca_eccv_002 |
| Trip.~+ Huber dist. | huber_distance_triplet_xy_000 |
| Log-ratio | ha0_lologratio_ma15_mi15_muTrue_renone_tu1_vl64_pca_eccv_002 |
| Multi-similarity | ha0_loms_loss_msTrue_muTrue_renone_tu1_vl64_pca_eccv_001 |
| Ours | al0.8_be15_ha0_lowms_ma15_mi15_msTrue_muTrue_renone_tu1_vl64_pca_eccv_000 |