My attempt at mixing my web dev knowledge with Machine Learning. Created a Neural Network Visualizer using streamlit 😍 and Flask

Trained a simple model with three layers on MNIST dataset and created an endpoint using Flask which first chosen a random image from MNIST dataset and runs the trained model against the choosed image and outputs the value of each neuron and the chosen image.

Used Streamlit to consume this API and display this data in an easy to understand and clean, usable frontend.

It took me more time to host this project on the free tier of Heroku then to make it, but it is now successfully hosted on Heroku. Check it out here. I also wrote a blog on how I created my application, consider checking it out here 😇

It might take a few second to load and generate the first output due to Heroku putting the containers to sleep after 30 minutes of inactivity

The files which have not been listed were used while deploying the project to Heroku and aren't of any importance while running the server locally

• build_model
  • Train_MNIST_model.ipynb: Used to train MNIST model and generate model.h5
• app.py: Contains the core logic of server and script which runs the model to generate output
• requirements.txt: Contains all the PyPi packages need to run this server

• app.py: Contains the core logic of frontend written using streamlit
• requirements.txt: Contains all the PyPi packages need to run this server

Make sure you have at least 2-3 GB of free RAM available because you will have to run two servers (Flask and streamlit) locally

Its a very simple procedure to run the project locally, It is just a matter of 5-6 lines of commands. It may take more time depending on your internet connection and if you already have the requirements or not (especially tensorflow 😜)

git clone https://github.com/jai-dewani/Visualize-Neural-Networks.git
cd Visualize-Neural-Networks

Open two terminals as you will have to keep two server up and running

cd backend
pip install requirements.txt
pyython app.py

Get the url in which your flask server is running like localhost:3000 or something.
Replace this value with the current placeholder URL in app.py in the frontend folder

pip install requirements.txt
streamlit run app.py