Handwritten Digit Classification :

This is a small deep learning project I've developed. It involves training a Convolutional Neural Network (CNN) on the MNIST dataset to accurately recognize handwritten digits. Subsequently, I established a web application using Flask to provide a user-friendly interface. This application includes a graphical user interface (GUI) allowing users to draw digits directly on a canvas. The model then processes these drawn digits to predict the corresponding numerical value.

To ensure a structured and organized development process, I adopted a modular coding approach. This approach involves breaking down the project into distinct steps, including data ingestion, data preprocessing, and model training. Additionally, I leveraged DVC (Data Version Control) to effectively track the entire pipeline, enhancing collaboration and reproducibility.

By following this systematic approach, I aimed to create an efficient and robust system for handwritten digit recognition, seamlessly integrating machine learning with web technology.

Project Screenshots

Setup This Project Locally

1. Create conda env or venv python == 3.9

2. Clone this repository

3. Install requirements.txt

pip install -r requirements.txt

4. Run Training Pipeline

dvc repro

This command will automaticaly run training pipeline. It will run all the steps in training

1. Ingest data: This step will download data from source url and create train.csv and test.csv

2. Training model: This step will create the model and do training and then save model in artifacts/models/model/h5

3. Evaluate model: This step will load trained model from directory and will evaluate trainig model.

5. Run web app

python app.py

visit http://127.0.0.1:3000/ in browser to access app

To track experiments

run command

mlflow ui

visit http://127.0.0.1:3000 in browser to see experiments.