AI-Digit Recognizer

This project is an implementation of a digit recognition model based on a Convolutional Neural Network (CNN) similar to ResNet, designed to participate in the Kaggle Digit Recognizer competition.

Overview

The goal of this project is to accurately classify handwritten digits from the popular MNIST dataset using a deep learning model. The model is built using a CNN architecture, inspired by ResNet, and achieves high accuracy through hyperparameter adjustments and training optimizations. See the competition here: https://www.kaggle.com/competitions/digit-recognizer/overview

Features:

• Convolutional Neural Network model inspired by ResNet.
• Learning rate adjustment on plateau for optimized training.
• Flexible hyperparameter tuning for device compatibility.
• Easy integration with Kaggle's Digit Recognizer competition dataset.

To Run the Script:

• To run, first download the datasets from the kaggle competition: https://www.kaggle.com/competitions/digit-recognizer/overview
• Adjust hyperparameters to work with your device, as well as data and test_data paths for the dataset
• Run the cnn_digits.py script in the terminal

Results:

98.467% accuracy