Here are the sample outputs from the model:

Welcome to my deployed computer vision project using YOLOv8 + Gradio, designed to detect doors and windows in architectural construction blueprints.

This application showcases a deep learning model trained on blueprint images to identify:

• 🚪 Doors
• 🪟 Windows

The goal was to automate detection in architectural layouts and assist with digitizing or verifying blueprint components.

• The app is powered by a custom-trained YOLOv8 model.
• You can upload a blueprint image via the Gradio interface.
• The model will return:
  • ✅ An annotated image showing detections
  • ✅ A JSON output with detection details in this format:

{
  "detections": [
    {"label": "door", "confidence": 0.91, "bbox": [x, y, w, h]},
    {"label": "window", "confidence": 0.84, "bbox": [x, y, w, h]}
  ]
}

1. Upload a construction blueprint image.
2. View the image with annotated bounding boxes.
3. Review the detection results in JSON format.

git clone https://github.com/Sagarkeshave/WinDoorDetection_YOLO.git
pip install -r requirements.txt
python app.py

Web App URL - https://huggingface.co/spaces/SagarKeshave/WinDoorDetection_YOLO

• Framework: Ultralytics YOLOv8
• Trained On: Custom blueprint dataset with annotated door and window classes.
• Performance: Optimized for fast inference on 2D plan layouts

This project demonstrates:

• Custom model training.
• Practical use of object detection in architecture
• Real-world deployment using Hugging Face Spaces
• Building user-friendly ML apps with Gradio

SAGAR G. KESHAVE
LinkedIn