MNIST Digit Classifier

A neural network built with Keras to classify handwritten digits from the MNIST dataset. The project includes comprehensive data visualization, training analysis with accuracy/loss metrics, and real image prediction capabilities. Features model training, evaluation, and inference for digit recognition with high accuracy.

Model Development

• Convolutional Neural Network (CNN) architecture
• Model training with validation split
• Hyperparameter tuning
• Model evaluation with accuracy and loss metrics

Data Analysis

• Data preprocessing and normalization
• Training history visualization
• Confusion matrix analysis
• Real-time prediction on custom images

Built with TensorFlow/Keras for deep learning model development. Utilizes NumPy and Pandas for data manipulation, and Matplotlib for comprehensive data visualization. Implements CNN architecture optimized for image classification tasks, achieving high accuracy on the MNIST digit recognition benchmark.