| language: en | |
| license: mit | |
| tags: | |
| - keras | |
| - mnist | |
| - image-classification | |
| - cnn | |
| datasets: mnist | |
| model-type: image-classification | |
| # MNIST Classification Model | |
| An improved CNN model for handwritten digit recognition, trained on the MNIST dataset. | |
| ## Model Architecture | |
| - Uses Convolutional layers (CNN) | |
| - Data Augmentation for improved performance | |
| - Batch Normalization | |
| - Dropout for preventing Overfitting | |
| - Dense layers with ReLU activation | |
| ## Parameters | |
| - Optimizer: Adam (lr=0.001) | |
| - Loss: Sparse Categorical Crossentropy | |
| - Metrics: Accuracy | |
| - Epochs: 20 (with Early Stopping) | |
| - Batch Size: 32 | |
| ## Performance | |
| Test Accuracy: 0.9884 | |