ML & DataBachelor's thesis · HAMK2024Completed
Smart Parking Assistant
Parking-space detection with a chatbot you can ask for a free spot.

- detection precision (YOLOv5l)
- 97.9%
- detection precision (YOLOv5l)
- chatbot intent recognition
- 92%
- chatbot intent recognition
Overview
My bachelor's thesis. A YOLOv5 model finds free and occupied parking spaces in camera footage, and a Rasa chatbot answers questions about availability in plain language. A Flask API handles the video processing, and Streamlit provides the interface.
Approach
- YOLOv5l fine-tuned on the PKLot dataset, with synthetic images added to cover more conditions.
- Rasa handles the conversation, for example "is there space near the entrance?"
- Video is processed locally, and only aggregated occupancy is stored.