The VIGOR dataset has been published in CVPR2021 (see paper). The original labels were based on an incorrect ground resolution for the aerial images. We corrected the labels of this dataset and the details can be found in Section A of the supplementary material of our paper (see supp).
The folder splits__corrected contains our corrected labels. After downloading the VIGOR dataset, the original splits
folder should be replaced by our splits__corrected
folder. The data should be organized as:
VIGOR
|_ Chicago
| |_ panorama
| |_ [prefix,latitude,longitude,].jpg
| |_ ...
| |_ satellite
| |_ [satelitte_latitude_longitude].jpg
| |_ ...
|_ NewYork
| |_ panorama
| |_ [prefix,latitude,longitude,].jpg
| |_ ...
| |_ satellite
| |_ [satelitte_latitude_longitude].jpg
| |_ ...
|_ SanFrancisco
| |_ panorama
| |_ [prefix,latitude,longitude,].jpg
| |_ ...
| |_ satellite
| |_ [satelitte_latitude_longitude].jpg
| |_ ...
|_ Seattle
| |_ panorama
| |_ [prefix,latitude,longitude,].jpg
| |_ ...
| |_ satellite
| |_ [satelitte_latitude_longitude].jpg
| |_ ...
|_ splits__corrected
The Jupyter notebook Visualize_Corrected_Labels can be used to visualize the corrected labels. The notebook assumes that the VIGOR dataset is placed in the datasets folder, i.e., the directory of the dataset should be DIR_TO_SLICEMATCH/SliceMatch/datasets/VIGOR
.