🐱🐶 Cats vs Dogs Image Classification using ResNet-50

This project applies deep learning and transfer learning techniques to classify images of cats and dogs. Leveraging the power of a pre-trained ResNet-50 model, the solution achieves high classification accuracy on the Kaggle Dogs vs. Cats dataset.

📎 View Full Notebook on Kaggle:
Cats & Dogs Image Classification | ResNet-50

---

📂 Project Structure

cats-dogs-resnet50/
├── data/                 # Directory for training and validation images
├── notebooks/
│   └── cats_dogs_resnet50.ipynb   # Main training and evaluation notebook
├── models/               # Saved model weights (optional)
├── outputs/              # Logs, plots, and predictions
├── requirements.txt      # Dependencies
└── README.md             # Project overview

---

🧠 Model Overview

• Architecture: ResNet-50 (pre-trained on ImageNet)
• Layers Modified: Top classification layer replaced with custom dense layer for binary classification
• Loss Function: Binary Crossentropy
• Optimizer: Adam
• Metrics: Accuracy

---

🧪 Key Steps

1. Data Preparation

   • Downloaded the Dogs vs. Cats dataset
   • Resized images to 224x224 to match ResNet-50 input format
   • Applied train-validation split and data augmentation

2. Model Building

   • Used Keras ResNet50 with include_top=False
   • Added custom fully-connected layers for binary classification
   • Compiled with Adam optimizer and binary_crossentropy loss

3. Training

   • Used early stopping and model checkpoint callbacks
   • Trained for multiple epochs with real-time data augmentation

4. Evaluation & Prediction

   • Evaluated on validation data
   • Visualized confusion matrix, classification report, and sample predictions

---

📊 Results

• Achieved ~98% validation accuracy
• Model generalizes well across different image types
• Visualizations confirm robust prediction across classes

---

📌 Requirements

• TensorFlow / Keras
• NumPy, Matplotlib, scikit-learn
• Jupyter Notebook
• PIL, tqdm

---

📈 Future Improvements

• Hyperparameter tuning
• Experiment with deeper architectures (e.g., EfficientNet, DenseNet)
• Deploy model via Flask or Streamlit

---

📎 References

• Kaggle Dataset: Dogs vs. Cats
• ResNet-50 Paper