DigiVis: Neural Network Vision for Digit Classification

A deep learning implementation for MNIST digit recognition using convolutional neural networks and computer vision techniques. This project combines modern neural architectures with advanced image processing for accurate digit classification.

Overview

DigiVis is a comprehensive implementation of various neural network architectures for digit recognition, utilizing the MNIST dataset. The project incorporates modern deep learning practices including data normalization, image transformations, and model evaluation metrics.

Features

• Multiple neural network architectures (CNN and Linear models)
• Data normalization and preprocessing
• Image transformations and augmentations
• Model training with performance metrics
• Weights & Biases integration for experiment tracking
• Comprehensive test suite
• CUDA support for GPU acceleration

Requirements

• Python 3.x
• PyTorch
• torchvision
• numpy
• pandas
• Pillow
• wandb
• matplotlib
• scikit-learn
• tqdm

Installation

1. Clone the repository

git clone https://github.com/Programmer-RD-AI/DigiVis.git

2. Install dependencies

pip install -r requirements.txt

Usage

Run the main training script:

python run.py

For interactive exploration, use the provided Jupyter notebook:

jupyter notebook test.ipynb

Model Configuration

• Image Size: 224x224
• Batch Size: 32
• CUDA enabled for GPU acceleration
• Random seed: 42 for reproducibility

License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.