Image Compression using K-Means Clustering and Convolutional Autoencoders

A project which performs compression and decompression of images from the MNIST dataset using autoencoder and K-Means clustering and deployment of the model using Flask.

Environment

• Coded in python 3.8
• Interactive python notebook editor - Google collab

Folders

• Notebooks - contains the google collab notebooks
• Flask App - contains the code required for the flask application
• Models - contains the pretrained models in h5 format
• Results - contains screenshots of the project outcome

Installation

To run the flask app in a windows environment

1. Install python 3.8
2. Run pip install virtualenv
3. Run mkdir project to create project directory
4. Run cd project to move to the project directory
5. Run virtualenv venv to create a virtual environment
6. Run .\Scripts\activate to activate the virtual environement
7. Run pip install tensorflow numpy Flask keras matplotlib pillow opencv-python scikit-image cv2 sklearn to install the dependencies
8. Copy the contents of the Flask app folder to your virtual environment and use command python app.py to run the app.

The python notebooks can be executed in Google collab by signing up for a Google account. Upload the pre-trained models into the collab runtime and load them to execute the code faster.