🐶 Dog Look-Alike Finder 🧑

Ever wondered which dog breed you resemble? This fun web application uses a deep learning model to analyze a photo of your face and find your canine twin from a dataset of over 100 dog breeds!

This project is a hands-on introduction to building a full-stack machine learning application, covering everything from data processing and feature extraction to building a web interface with Flask.

✨ Features

• Simple Web Interface: Easy-to-use webpage for uploading your photo.
• AI-Powered Face Detection: Uses OpenCV to automatically detect and crop the human face from the uploaded image, ensuring more accurate comparisons.
• Deep Learning Feature Extraction: Employs a powerful pre-trained Convolutional Neural Network (VGG16) to understand the unique visual features of a face.
• Efficient Similarity Search: Compares facial features using Cosine Similarity to find the best match quickly.

---

🔧 How It Works

The application's logic is broken down into a simple but powerful machine learning pipeline:

1. Data Pre-processing (One-time step): The process_dataset.py script first runs through the entire Stanford Dogs Dataset. For each dog image, it performs a center-crop and uses the VGG16 model to extract a feature vector (a numerical representation of the image). All these vectors are saved to a single file, dog_features.pkl, for fast lookups later.

2. User Upload: A user uploads their photo via the Flask web application.

3. Human Face Detection: The backend uses an OpenCV Haar Cascade classifier to find the user's face in the uploaded photo and crops the image to that region.

4. Feature Extraction: The cropped face image is then passed to the same VGG16 model to generate its unique feature vector.

5. Comparison: The application calculates the Cosine Similarity between the user's feature vector and every dog vector loaded from dog_features.pkl. The dog with the highest similarity score (lowest cosine distance) is the winner!

6. Display Results: The original photo and the winning dog photo are displayed to the user.

---

🛠️ Technology Stack

• Backend: Python, Flask
• Machine Learning: TensorFlow, Keras
• Computer Vision: OpenCV, Pillow
• Scientific Computing: NumPy, SciPy
• Frontend: HTML, CSS

---

🚀 Setup and Installation

Follow these steps to get the project running on your local machine.

Prerequisites

• Python 3.8+
• pip (Python package installer)

1. Clone the Repository

git clone https://github.com/jasperlawrencee/watda-dog-r-u.git
cd watda-dog-r-u

2. Create a Virtual Environment

It's highly recommended to use a virtual environment to keep project dependencies isolated.

# For macOS/Linux
python3 -m venv venv
source venv/bin/activate

# For Windows
python -m venv venv
.\venv\Scripts\activate

3. Install Dependencies

Install all the required Python libraries from the requirements.txt file.

pip install -r requirements.txt

4. Download Required Data

• Stanford Dogs Dataset:

  • Download the dataset from Kaggle.
  • Unzip the file and place the images/Images folder into the root of the project directory. Make sure the folder is named Images.

• Haar Cascade Classifier:

  • Download the haarcascade_frontalface_default.xml file from OpenCV's GitHub repository.
  • Place this file in the root of the project directory.

5. Pre-process the Dog Dataset

This is a crucial one-time step. Run the processing script to generate the feature vectors for the entire dog dataset. This will create the dog_features.pkl file.

Note: This process can take 15-30 minutes depending on your computer's performance.

python process_dataset.py

---

▶️ How to Run the Application

Once the setup is complete, you can start the Flask web server.

flask --app app run

Open your web browser and navigate to: http://127.0.0.1:5000

You should now see the application's homepage. Upload a photo and enjoy!

---

📁 Project Structure

.
├── app.py                                  # Main Flask application
├── feature_extractor.py                    # Core ML/CV functions (cropping, feature extraction)
├── process_dataset.py                      # Script to pre-process the dog dataset
├── requirements.txt                        # Project dependencies
├── haarcascade_frontalface_default.xml     # OpenCV face detector
├── dog_features.pkl                        # Generated file of dog feature vectors
├── Images/                                 # Folder for the Stanford Dogs Dataset
│   └── n02085620-Chihuahua/
│   └── ...
├── templates/                              # Flask HTML templates
│   ├── index.html
│   └── result.html
└── uploads/                                # Stores user-uploaded images temporarily

---

🙏 Acknowledgements

• This project uses the Stanford Dogs Dataset.
• Feature extraction is powered by the VGG16 model, pre-trained on ImageNet.
• Face detection is performed using OpenCV.