gabriellidenor/svhn-digit-recognition
🔢 Street View House Numbers (SVHN) Digit Recognition Pipeline This repository hosts the dataset layer and acts as an integration hub for a modular, production-ready machine learning pipeline. The project is designed to classify cropped house numbers from street-view imagery using deep learning architectures. 📂 Project Structure svhn_digit_recognition/ │ ├── data/ │ ├── raw/ # Original, immutable data dump (.h5 file) │ └── processed/… See the full description on the dataset page: https://huggingface.co/datasets/gabriellidenor/svhn-digit-recognition.
🔢 Street View House Numbers (SVHN) Digit Recognition Pipeline
This repository hosts the dataset layer and acts as an integration hub for a modular, production-ready machine learning pipeline. The project is designed to classify cropped house numbers from street-view imagery using deep learning architectures.
📂 Project Structure
svhn_digit_recognition/
│
├── data/
│ ├── raw/ # Original, immutable data dump (.h5 file)
│ └── processed/ # Final canonical datasets used for modeling
│
├── notebooks/
│ └── exploration.ipynb # Kept for data visualization & initial testing
│
├── src/ # Main source code directory
│ ├── __init__.py
│ ├── data_loader.py # Scripts to read, split, and normalize images
│ ├── models.py # Houses Neural Network and CNN model definitions
│ └── evaluate.py # Classification reports, confusion matrices, metrics
│
├── config.py # Centralized hyperparameters (epochs, batch size, paths)
├── train.py # Main pipeline orchestrator to run the experiment
└── requirements.txt # Package dependencies (TensorFlow, scikit-learn, h5py)