/soi-space-ds

Summer of Innovation 2022 Space Data Science

Primary LanguagePythonMIT LicenseMIT

Summer of Innovation 2022 Space Data Science

This software is shared under MIT License.

Team Kepler

This is a standalone web-app used for classification of planetary data. The web-app can be setup by following the shown steps.

Prerequisites

  1. Linux system.
  2. Python / Python3 installed.
  3. Latest version of pip installed.

Steps

  1. Open your bash terminal.
  2. Clone this repository in any directory of your choice.
git clone https://github.com/shashankp28/soi-space-ds.git
  1. Run the following command to move into the cloned repository.
cd soi-space-ds
  1. Run one of these commands to install tkinter, depending upon your system's python configuration (python / python3).
sudo apt-get install python3-tk

sudo apt-get install python-tk
  1. Execute the setup.sh file to install dependencies and create run.sh file.
chmod 755 setup.sh && ./setup.sh

When prompted, choose whether to install packages on a virtual environment. yes recommended

  1. Host the web-app locally using the following command
./run.sh

Once the setup is complete, the web-app can be opened using loalhost Port 8501. Use Ctrl+C inside the terminal to stop.

Important

Once the app is setup, you can host the web-app using only step 5.

Using the application

  1. Initially upload a csv file in the format shown in the web-app.
  2. Next navigate to the Docs tab from the side nav bar.
  3. Detailed instructions on using the application is given, including a short video.

Note

  • The notebook and data used for training can be found under the following directories:

    1. ML/SDS_MODEL.ipynb
    2. ML/data_full.csv

  • Documentation for ML model is named as Documentation_Kepler.pdf

  • Documentation for the Web-App can be found under the Docs tab of the Web-App itself.

  • The predictions are present in the predicted coloumn in downloaded files.

The Random Forest Model couldn't be incorporated as it's size was around 3.5 GB and would not be feasible for a stand-alone application. The model can be run on Google Colab.