SVHN Digit Recognition 🏘️🔢

🚀 Project Overview

Classify street-level digits using deep learning! This project uses a subset of the SVHN dataset (600k+ labeled images) to experiment with Feedforward Neural Networks (FNNs) and Convolutional Neural Networks (CNNs), aiming to find the most accurate model for digit recognition.

📊 Highlights

Dataset: .h5 format, 32×32 RGB images

Models Tested: FNNs & CNNs

Results: CNNs outperform FNNs with high accuracy

Visuals: Prediction vs. Ground Truth comparisons

🛠️ Tech Stack

Python 3.x | TensorFlow / Keras

NumPy, Pandas | Matplotlib / Seaborn

h5py for dataset handling

🎯 Quick Start git clone https://github.com/rishovm/Pixel2Number.git cd svhn-digit-recognition pip install -r requirements.txt python train_model.py python evaluate_model.py

🔮 Future Work

Train on full SVHN dataset for max performance

Use data augmentation for robustness

Explore ResNets or Transformers for digits

✨ Author

Dr. Rishov Mukhopadhyay