**Background**
As public transport moves towards driverless vehicles, in-cabin cameras are proposed to detect critical passenger events such as falls, medical emergencies and conflicts. However, continuous monitoring of passengers raises GDPR concerns, since raw footage is personal data and is costly to justify, store and protect under the regulations data minimization principle.


To reduce the collection of personal data, passengers could be represented as skeleton points (pose estimation data) instead of raw video. This thesis examines whether skeleton data alone is sufficient to reliably detect passenger events and behaviors, as a way of making GDPR compliance easier to achieve.


**Work Description**

- Build a pipeline that extracts skeleton sequences from cabin video using an existing pose estimation model.
- Design and train a classifier that detects relevant passenger events from skeleton sequences alone.
- Collect or reuse labelled scenario data (open datasets and/or scripted recordings) reflecting realistic bus-cabin conditions.
- Benchmark detection accuracy and real-time latency against other detection models.
- Assess, using GDPR's data-minimization principle, what personal data the skeleton representation avoids collecting compared to conventional video.

**We think the thesis will contain the following parts,**

- Literature review on pose-based activity/fall recognition and privacy-preserving video analytics.
- Design and implementation of the skeleton-extraction and event-classification pipeline.
- Empirical evaluation of accuracy, latency, and privacy exposure versus a conventional video based approach.
- Recommendations for how detection performance can be balanced against data minimization in this type of system.

**Qualifications**

- Programming experience in Python.
- Understanding of machine learning and image processing techniques.
- Background in Computer Science, Data Science AI or equivalent.
- Good knowledge in Swedish and English, both in writing and speeking.

*Tags: Computer Vision, Pose Estimation, Privacy Preserving AI, Autonomous Vehicles, GDPR*
To give you the best possible support during your thesis, we’d like you to be able to come to the office connected to the project and spend most of your time working from there.


**Application:**
We look forward to receiving your resume, and preferably, a personal letter in which you explain why you want to write your thesis with Syntronic.


We screen and evaluate applications on an ongoing basis. The thesis project may be filled before the application deadline.