Thesis Proposal - Pose based occupant monitoring for driverless buses: Privacy preserving detection of critical passenger events

Driverless buses need ways to identify passenger incidents such as falls, medical emergencies and conflicts without unnecessarily collecting personal data. This thesis investigates whether skeletal pose sequences can support reliable event detection instead of retaining conventional passenger video. The aim is to evaluate detection performance alongside privacy benefits under GDPR’s data-minimization principle.

The assignment

• Review research on pose-based activity and fall recognition and privacy-preserving video analysis.
• Implement a pipeline that converts cabin footage into skeleton sequences using an existing pose-estimation model.
• Develop and train a classifier to recognize passenger events using only the extracted skeleton sequences.
• Gather or reuse labelled scenarios from public datasets and/or staged recordings that represent realistic conditions inside buses.
• Compare detection accuracy, real-time latency and privacy exposure with conventional video-based methods and other detection models.
• Identify which personal data skeleton representations avoid collecting and recommend ways to balance detection quality with data minimization.

Who we are looking for

• Academic background in computer science, data science, artificial intelligence or a related discipline.
• Experience programming in Python.
• Knowledge of machine learning and image processing.
• Good spoken and written Swedish and English.

What we offer

• Support during the thesis, with most work expected to take place at the office associated with the project.

Practical details

Location: Linköping. Applicants should submit a CV, preferably accompanied by a personal letter explaining their interest in completing a thesis at Syntronic. Applications are reviewed continuously, and the position may be filled before the application deadline.