Computation | Visualization | Vectorization
This code accompanies the paper "A Survey of Vectorization Methods in Topological Data Analysis, Dashti Ali, Aras Asaad, Maria Jose Jimenez, Vidit Nanda, Eduardo Paluzo-Hidalgo, and Manuel Soriano-Trigueros"
This web application can be used to compute and visualize persistence barcodes as well as the thirteen vectorization techniques investigated in this paper. Beside running online, the companion web application can also be downloaded and run on users local machines to compute and visualize barcodes from large point-clouds and images. This no code application can be used as a convenient tool to explore different vectorization techniques by beginners and experts alike. The WebApp can be accessed via the following link:
https://persistent-homology.streamlit.app
The app is implemented in Python using Streamlit library together with other Python libraries. Streamlit is a Python framework for creating web applications for machine learning and data science projects. The application has a sidebar to choose between different types of input data and displays several options for data visualization and exportation. A sample image/point-cloud from each of the datasets used in this paper is preloaded into the web application together with their associated vectorization techniques and visualizations. The user can either select from one of these data samples or upload an image or point cloud file for which the different vectorization methods will be computed and visualized. Uploaded images by users will be resized to 64-rows by 64-columns and converted to gray-scale. The user can only upload 3D point cloud and only processing the first 100 points for their uploaded 3D point cloud. These restrictions are removed when the user run the application on their local machines. Furthermore, cubical complexes and Vietoris-Rips filtrations are used by default to build persistence barcodes for images and point clouds uploaded by the user,respectively.
The app will compute the persistence barcodes of dimension 0 and 1 and the user can select to have them plotted by ticking a check box (persistence diagram representation can also be displayed). All barcode vectorization methods considered in this paper can be computed and visualized in different formats (tables, graphics, scattered plots), depending on the type of vectorizarion. Finally, an export button is associated with each barcode/diagram and vectorization method so that users can easily download associated data for further data analysis tasks.
To run the app locally on your machine, first clone or download this repository. If you download the repository as zip file, make sure that you have extracted the content of the zip file into a folder. Then follow these steps to run the app on your local machine:
- Make sure you have python (version >= 3.8 & < 3.11) installed on you machine
- Open command window in windows or terminal in Mac/Linux.
- Navigate to the folder where you cloned the repository or extracted the content of the downloaded zip file.
- Execute the following command to install
streamlitpackage:
pip install streamlit
- All dependencies for the app to run are located in requirements.txt file, to install all these dependencies in one go execute this command:
pip install -r requirements.txt
- If you have added python Scripts folder to your system PATH execute the following command to run the app:
streamlit run app.py local
- If step 6 does not work for you, then execute the following command to run the app:
python -m streamlit run app.py local