This is a collection of scripts to cluster UI layouts produced by Rico.
Convert the view hierarchies from XML to the JSON format used by Rico.
python convert_view_hierarchy_format.py <absolute_path_to_data_folder>
Represent each UI with an image with three separate segments for text elements, icons and images.
python create_ui_representation.py <absolute_path_to_data_folder>
Combine these images into one numpy array that can be used for training an autoencoder.
python create_ae_inputs.py
Train the autoencoder.
python train_ae.py
Reconstruct the input images from the compressed representation produced by the autoencoder. Output images will be located in the reconstructed_imgs folder for inspection.
python visualize_ae_reconstruction.py
Run k-means clustering on the encoded images for the images.
python run_kmeans.py
Produce HTML files to visualize the UI clusters. HTML files will be located in cluster_viz folder for inspection.
python create_uv_map.py python visualize_clusters.py <absolute_path_to_data_folder>