🧠 Deep Face Detection with Bounding Box Regression

This project implements a deep learning pipeline using TensorFlow and VGG16 to detect faces and predict bounding boxes from webcam input. It features real-time predictions with custom-trained models using image augmentation and bounding box regression.

---

🧩 Project Highlights

• 📸 Real-time face detection from webcam
• 🔍 Bounding box prediction (regression)
• 🧠 VGG16-based feature extraction
• 🧪 Custom training pipeline with augmentation
• 📁 JSON label handling
• 📊 TensorBoard support

---

🗂 Folder Structure

Deep_Face_Detecton_Model/
├── data/               # Original images and labels
├── aug_data/           # Augmented training/test/val data
├── logs/               # TensorBoard logs
├── tfenv/              # Python virtual environment (ignored)
├── main.py             # Main training, augmentation, and detection script
├── test.py             # Image collection via OpenCV
├── facetracker.h5      # Saved model file

---

▶️ How to Run

1. Install dependencies:

pip install -r requirements.txt

2. Collect images via webcam:

python test.py

3. Train the model:

python main.py

4. Real-time detection after training:

python main.py  # Last section activates webcam

---

💡 Notes

• Model is saved as facetracker.h5
• Uses Albumentations for image augmentation
• Bounding boxes are normalized and rescaled during inference

---

📜 License

For research and educational use only.