A computer vision-based X-ray object detection app prototype that identifies potentially dangerous items in X-ray scans. Built with a YOLOv11 custom-trained model and powered by a Python-based backend (Flask) and frontend (Streamlit).

• ✅ Real-time object detection on X-ray images
• 🧠 Custom-trained YOLOv11 model
• 📊 Detection confidence & status reports
• 📸 Annotated output images with bounding boxes
• 🗃️ Upload your own image or use sample images
• 💾 Export detection logs as CSV

1. User uploads or selects an image through the Streamlit frontend.
2. Image is sent to the Flask API endpoint at /predict.
3. The YOLOv11 model performs inference on the image.
4. The API returns detected object data and the annotated image.
5. The frontend displays the detection results and allows CSV export of the detected objects.

• Model: YOLOv11 (by Ultralytics)
• Backend: Flask (for handling prediction requests)
• Frontend: Streamlit (for the web interface)
• Database: PostgreSQL (for logging detection history)
• Image Handling: OpenCV

• Upload image files and detect multiple objects.
• View annotated image with bounding boxes.
• Download detection results as CSV.
• (Optional) Log detection history to PostgreSQL.

Data Source: https://universe.roboflow.com/siewchinyip-outlook-my/sixray

• Trained for 32 epochs on Yolov11m.pt base
• Metrics:
  • Model accuracy measured on validation set
  • mAP50: 0.906
  • mAP50-95: 0.643
  • Precision: 0.92
  • Recall: 0.82

python app.py
streamlit run app_frontend.py