CNN Project – Stanford Dogs Dataset

This project applies Convolutional Neural Networks (CNNs) with transfer learning using EfficientNetB7 to classify dog breeds from the Stanford Dogs dataset. The dataset contains 120 dog breeds, making it a fine-grained image classification task.

---

📂 Contents

• Stanford_Dogs.ipynb – Main Jupyter Notebook with model training and evaluation
• requirements.txt – List of dependencies
• training_log.txt – Accuracy and loss logs

---

⚙️ Requirements

• Python 3.x
• TensorFlow / Keras
• NumPy
• Matplotlib
• scikit-learn

Install dependencies:

pip install -r requirements.txt

---

🚀 How to Run

1. Download the Stanford Dogs dataset via TensorFlow Datasets:

import tensorflow_datasets as tfds
ds, info = tfds.load("stanford_dogs", with_info=True, as_supervised=True)

2. Place the dataset in a folder named data/ (or adjust the path in the notebook).

3. Run the notebook:

jupyter notebook Stanford_Dogs.ipynb

---

📊 Results

Model: EfficientNetB7 (Transfer Learning)

• Training Accuracy: 92.76%
• Validation Accuracy: 95.64%
• Test Accuracy: 95.64%

---

📝 Notes

• This project demonstrates the effectiveness of EfficientNetB7 for fine-grained image classification.

• Future improvements could include:

  • More advanced data augmentation
  • Fine-tuning additional layers
  • Trying Vision Transformers (ViTs) or other modern architectures