🐱 Cat Classifier from Scratch (No ML Libraries)

This project implements a deep neural network from scratch using NumPy to classify images as cat vs non-cat.

---

📌 Features

• Forward and backward propagation manually implemented
• Supports ReLU, Sigmoid, and Tanh activations
• Trains on catvnoncat.h5 dataset from Andrew Ng's DL course
• Visualizes misclassified test examples
• Can predict on custom user-supplied image

---

🧠 Model Architecture

• Input layer: 12288 (64×64×3)
• Hidden layers: Configurable
• Output: Sigmoid activation for binary classification

---

🚀 Training Results

• Train Accuracy: 98%
• Test Accuracy: 70–75%
• Optimized using full-batch gradient descent
• Cost plotted to show convergence

---

🖼️ Example Output