Dog Vision - Dog Breed Image Classifier

A deep learning project that uses computer vision to identify dog breeds from images using TensorFlow and transfer learning.

🐕 Overview

This project implements an end-to-end multiclass image classifier capable of identifying 120 different dog breeds from photographs. Built using TensorFlow and Keras with transfer learning techniques, the model achieves 68% accuracy across all breed categories.

🎯 Features

• Multi-class Classification: Identifies 120 different dog breeds
• Transfer Learning: Leverages pre-trained MobileNetV2 architecture for efficient training
• Data Augmentation: Implements advanced preprocessing techniques for improved model robustness
• Performance Optimization: Uses TensorBoard callbacks to reduce training time by 50%+
• End-to-end Pipeline: Automated data preprocessing from raw images to normalized tensors

🛠️ Technologies Used

• Python: Core programming language
• TensorFlow/Keras: Deep learning framework for model building and training
• Scikit-Learn: Machine learning utilities and metrics
• Pandas: Data manipulation and analysis
• Matplotlib: Data visualization and result plotting
• Google Colab: Development and training environment
• NumPy: Numerical computing operations

🏗️ Architecture

The model uses transfer learning with MobileNetV2 as the base architecture:

• Base Model: MobileNetV2 (pre-trained on ImageNet)
• Custom Layers: Additional dense layers for dog breed classification
• Input Shape: 224x224x3 RGB images
• Output: 120 classes (dog breeds)
• Optimization: Adam optimizer with custom learning rate scheduling

📊 Model Performance

• Accuracy: 68% on test dataset
• Training Data: 10,000+ dog images across 120 breeds
• Training Time: Optimized using TensorBoard callbacks (50% reduction)
• Validation: Comprehensive evaluation with confusion matrices and classification reports

🚀 Getting Started

Prerequisites

pip install tensorflow
pip install scikit-learn
pip install pandas
pip install matplotlib
pip install numpy

Installation

1. Clone the repository:

git clone https://github.com/nryee2005/Dog_Vision.git
cd Dog_Vision

2. Install required dependencies:

pip install -r requirements.txt

3. Open the Jupyter notebook or run in Google Colab for best performance.

Usage

1. Data Preparation: The notebook includes automated preprocessing pipeline
2. Model Training: Run the training cells to fine-tune the MobileNetV2 model
3. Evaluation: Analyze model performance using built-in metrics and visualizations
4. Prediction: Use the trained model to classify new dog images

📁 Project Structure

Dog_Vision/
├──Dog Vision/
   ├── train/                                          # Train data set
   ├── test/                                           # Test data set
   ├── models/                                         # Saved models 
   ├── logs/                                           # TensorBoard logs
   ├── custom images/                                  # Custom images used
   └── full_model_predictions_1_mobilenetV2.csv        # Predictions on test dataset using full model
├── dog_vision.ipynb                                # Jupyter notebook
├── requirements.txt                                # Project dependencies
└── README.md                                       # Project documentation

🎓 Learning Outcomes

This project demonstrates proficiency in:

• Deep learning and neural networks
• Transfer learning techniques
• Computer vision applications
• Data preprocessing and augmentation
• Model optimization and performance tuning
• TensorFlow/Keras framework usage

👨‍💻 Author

Nathan Yee

• Email: ryee.nathan@gmail.com
• LinkedIn: nathan-r-yee
• GitHub: nryee2005

---