6G Platform for AI Computer Vision

Important results from the project

The goal of developing AI-driven methods for smarter door systems was met. The project established an automated data collection pipeline and a unique dataset for pedestrian behavior. Evaluations showed that geometric data is often sufficient for reliable short-term predictions, while pose features can stabilize specific models. The solution was validated on embedded hardware, confirming the feasibility of energy-efficient Edge AI solutions.

Expected long term effects

In the long term, the results are expected to contribute to reduced energy consumption in public buildings by minimizing unnecessary door openings. The project strengthens knowledge regarding 6G-connected IoT devices and distributed AI, specifically concerning the optimization of computer vision on embedded systems. The methodology for privacy-secure data collection also sets a standard for the future development of ethical AI in public environments.

Approach and implementation

The project was carried out as a degree project in collaboration between RISE and Assa Abloy. The work began with a small-scale data collection in a home environment to optimize parameters, followed by a large-scale collection in an office environment. Machine learning models (e.g. LSTM, CNN) were trained and evaluated systematically. Despite varying data complexity, the automatic annotation worked well and the schedule was met.