This project is a Convolutional Neural Network (CNN) model that classifies handwritten digits, which training was based on the MNIST dataset. The purpose of this project is to demonstrate the capabilities of deep learning techniques for image classification tasks, specifically focusing on recognizing handwritten numbers.

The MNIST dataset consists of 70,000 grayscale images of handwritten digits (0-9). This project employs a stacked CNN to accurately classify these images. The motivation behind this project is to showcase how deep learning can be effectively utilized for image recognition tasks, enhancing understanding of neural networks and their applications.

You can checkout the model demo here.

TensorFlow

TensorFlow Datasets (MNIST)

Keras

NumPy

Pillow (Image processing)

Flask (Backend/Templating)

Streamlit (Deploy/Frontend)

Image Preprocessing/Augmentation: Generalize and prepare images for better model performance.

Model Outline: Outline the model architecture, with number of layers and nodes, defining activation functions and so.

Training and Evaluation: Trains the CNN model and evaluates its accuracy on the test dataset.

Frontend for Real-Time Prediction: Allows users to draw on a canvas and receive predictions.