Handwritten Number Prediction

The project includes an automated dataset generation pipeline that extracts image features and creates a structured CSV dataset containing 784 pixel features for each image along with its corresponding digit label. A Scikit-learn MLPClassifier with two hidden layers is then trained on the generated dataset to learn patterns associated with handwritten digits from 0 to 9. For prediction, a new image is processed using the same preprocessing pipeline and converted into 784 normalized pixel values before being passed to the trained neural network. The predicted digit is then returned as the final output. The project demonstrates practical implementation of image preprocessing, dataset generation, feature extraction, neural-network classification, and image-based prediction using Python.