Overview §
Built a deep learning model to classify handwritten digits (0–9) using Python and TensorFlow/Keras.
Tested multiple architectures (Dense Neural Networks, Convolutional Neural Networks) and achieved the best results with a CNN.
Techniques Used §
- 🏗️ Clean CNN architecture with convolutional, pooling, and fully connected layers.
- 📊 Data preprocessing: Normalized and reshaped images for efficient training.
- 🤖 Model evaluation: Monitored accuracy, confusion matrix, and training/validation curves.
- ⚡ High performance: Achieved over 96.8% test accuracy with fast training and inference.
- ✅ Reliable predictions: Correctly classified most handwritten digits with high confidence.
Results §
- 🎯 Achieved over 96.8% test accuracy with the CNN model.
- ⚡ Fast training and inference, enabling efficient experimentation.
- ✅ High-confidence predictions for most handwritten digits.
- 🌟 Demonstrated robust generalization across unseen test images.