MNIST 1‑9 Digit Classifier (Dense‑only)

Fully‑connected neural network (no CNN) that classifies hand‑written digits 1‑9 from grayscale MNIST images.

Item	Details
Dataset	MNIST – filtered to digits 1‑9 (60 000 train / 10 000 test → 54 000 / 9 000 after filtering)
Input	28×28 pixels → flattened to 784
Architecture	Dense(256 ReLU) → Dropout(0.3) → Dense(128 ReLU) → Dropout(0.3) → Dense(9 Softmax)
Optimizer	Adam
Accuracy	⬜ 0.96

---

Highlights

• Pure MLP – demonstrates that even without convolutions, a dense network can classify digits reasonably well.
• Data Filtering – removed digit 0 to create a 9‑class problem (1‑9).
• Visualization – notebook shows sample inputs and highlights first 10 mis‑classifications in red.

---

How to Run

pip install tensorflow matplotlib scikit-learn
python mnist_dense_1to9.ipynb   # or open in Google Colab