Make sure the Hloc repository and this repository are in the same directory. Make sure to install Hloc with all its dependencies. https://github.com/cvg/Hierarchical-Localization
Download the dataset into datasets/ and unzip it:
!wget http://cvg.ethz.ch/research/local-feature-evaluation/South-Building.zip -P datasets/
!unzip -q datasets/South-Building.zip -d datasets/
Put the file ground_truth_adjusted.txt into the South-Building folder.
The overall method needs masks. The masks are stored as pkl file which can be generated from the segmentation model. You can download the weights that have been used from here: weights link. Once the masks are generated, put the masks folder into the South-Building path.
Pixellib BatchNormalization may not always show expected behaviour. Add the following line to the site package for semantic labelling incase the issue occurs: ''' from tensorflow.keras.layers import BatchNormalization '''
Open the Jupyter Notebook to run the code.