A Convolutional Neural Network (CNN) to classify handwritten digits from the MNIST dataset with a test accuracy around 98-99%. The model was built using TensorFlow and Keras, using the tensorflow.keras.datasets.mnist package to load and preprocess the data. This project demonstrates model training, evaluation, and performance visualization through confusion matrices and prediction plots.
- Dataset: MNIST dataset containing 70,000 grayscale images of handwritten digits at 28x28 pixels.
- Model: CNN model with Conv2D, MaxPooling2D, Flatten, and Dense layers for digit classification.
- Tools Used: TensorFlow, Keras, NumPy, Matplotlib, Seaborn, and Scikit-learn.
- Results: Achieved a test accuracy around 98-99% and generated comprehensive performance metrics including a confusion matrix and classification report.