PytorchMNIST-CNN

This is an implementation of a CNN (Convolutional neural network) to classify hand written digits from the MNIST dataset in pytorch

This project also contains code to serve the model in simple API built using flask and python

This project was made for learning purposes and to explore CNN's and image classification, I am only learning so there things that can be done in a better way please feel free to leave an issue with any suggestions

Made Using

• pytorch for Machine Learning
• Flask for the backend
• PIL (Python Imaging Library) for reading in images

Resources and inspiration

In order to create this application I followed along with

• This Tutorial on YouTube by FreeCodeCamp about Deep learning in pytorch
• This Tutorial on YouTube by Python Engineer about deploying a pytorch model with a flask API

How to setup the project

In order to use this project locally you will need to have python installed, which you can download from Here

Installation

After downloading and installing python you need to set up a virtual environment on your machine

For windows users:

python -m venv [Name]

For Mac users:

python3 -m venv [Name] 

where Name is the name of the environment

Unfortunately I haven't used a linux system 😢

Go into the environment

This is same for Mac and Windows

cd [Name]

Inside this directory there should be a folder called "venv"

the next step is to clone the project from GitHub

Running the program

In order to run the program activate the virtual environment, you can activate the environment using

Windows:

venv\Scripts\activate.bat

Mac:

source venv/bin/activate

your terminal or command prompt should show a

(venv) 

in front

then you need to install the dependencies for the project

Windows:

pip install -r requirements.txt

Mac:

pip3 install -r requirements.txt

now you can run the script

Windows:

cd app
run.bat

run.bat is a small script to set the environment variables for flask and start up the backend run.sh is same as run.bat for mac

On Mac this process is a tiny bit more complicated

cd app
chmod +x run.sh
./run.sh

and a server should be up and listening on localhost port 5000

todo

• Accept base64 images
• Add a some front end for the project
• Deploy it on either Heroku or Azure (TBD)