Real-Time Object Recognition (Raspberry Pi 5)
PwC Emerging Technologies Lab
- Built an AI-powered real-time object recognition system on Raspberry Pi 5, overcoming hardware limitations to deliver a responsive user experience.
- Implemented YOLOv11, trained on the Objects365 dataset, enabling detection of over 2 million object classes for a variety of use cases ranging from road safety to interior design.
- Custom compiled OpenCV from source in Python to optimise performance for the Pi environment.
- Integrated hardware, including a Samsung USB webcam for live video input, a custom compact 3D-printed case for deployment, and an external monitor for live video output.
- Designed a live UI overlay showing detected objects with confidence percentages, making results interpretable in real time.
- Managed code and version control on GitHub, maintaining a clear commit history and documentation for reproducibility.