Viewpoint Invariant Dense Matching for Visual Geolocalization: PyTorch implementation

This is the implementation of the ICCV21 paper:

G Berton, C. Masone, V. Paolicelli and B. Caputo, Viewpoint Invariant Dense Matching for Visual Geolocalization

Setup

First download the baseline models which have been trained following the training procedure in the NetVLAD paper. We provide a script to download the six models used, which are a combination of 3 backbone encoders (AlexNet, VGG-16 and ResNet-50) with 2 pooling/aggregation layers (GeM and NetVLAD).

python download_pretrained_baselines.py

Then you should prepare your geo-localization dataset, so that the directory tree is as such:

dataset_name
└── images
    ├── train
    │   ├── gallery
    │   └── queries
    ├── val
    │   ├── gallery
    │   └── queries
    └── test
        ├── gallery
        └── queries

and the images are named as @UTM east@UTM north@[email protected]

Dependencies

See requirements.txt

Training

You can train the model using the train.py, here's an example with the lightest/fastest model (i.e. AlexNet + GeM):

python train.py --arch alexnet --pooling gem --resume_fe pretrained_baselines/alexnet_gem.pth

For a full set of options, run python train.py -h. The script will create a folder under ./runs/default/YYYY-MM-DD_HH-mm-ss where logs and checkpoints will be saved.

Evaluation

Coming soon.

BibTeX

If you use this code in your project, please cite us using:

@InProceedings{Berton_ICCV_2021,
    author    = {Berton, Gabriele and Masone, Carlo and Paolicelli, Valerio and Caputo, Barbara},
    title     = {Viewpoint Invariant Dense Matching for Visual Geolocalization},
    booktitle = ICCV,
    month     = {October},
    year      = {2021},
    pages     = {12169-12178}
}