🧠 Counting Objects and Face Detection using OpenCV

This project uses YOLOv8 and OpenCV to detect and count objects, specifically focusing on identifying people and then detecting faces within the detected persons. It is implemented in Google Colab and leverages both Ultralytics YOLO and Haar Cascades.

🔍 Features

• Upload and analyze an image using Google Colab.
• Detect all objects using the YOLOv8 model.
• Highlight and count the number of persons detected.
• Perform face detection on each person using OpenCV's Haar cascade classifier.
• Save the annotated image with bounding boxes and labels.
• Display the final output image in the notebook.

📸 Demo

🛠️ Technologies Used

• Python
• OpenCV
• Ultralytics YOLOv8 (yolov8n.pt)
• Google Colab
• Haar Cascade Classifier

📦 Requirements

To run this project in Google Colab:

pip install ultralytics opencv-python

No manual installation of cv2 is needed in Colab — it is preinstalled.

📁 Project Structure

.
├── counting_objects_face_detection.ipynb
├── output_with_boxes.jpg
└── README.md

🚀 How It Works

1. Upload an image in Google Colab.
2. YOLOv8 detects all objects in the image.
3. Each detected object is labeled and counted.
4. If a detected object is a person, it is passed to the Haar Cascade classifier.
5. Faces are detected inside the person region.
6. Output is saved as output_with_boxes.jpg with all detections highlighted.

🔎 Sample Code Snippet

model = YOLO("yolov8n.pt")
results = model(image)
detections = results[0]
total_objects = len(detections.boxes)

# Face detection on detected 'person' boxes
face_cascade = cv2.CascadeClassifier(cv2.data.haarcascades + 'haarcascade_frontalface_default.xml')

💡 Future Improvements

• Switch to deep learning-based face detection for better accuracy.
• Add support for real-time webcam detection.
• Integrate a simple web UI using Gradio or Streamlit.
• Count and visualize other specific objects (e.g., vehicles, animals).

👤 Author

Deepak Misal 🔗 LinkedIn Profile

Samarth Bhutnal 🔗 LinkedIn Profile

---

Let me know if you'd like a version with Colab badges or want to auto-generate the sample image from your code output.