Master thesis: Multi-view object detection - architectures and training methods

Transport is at the core of modern society. Imagine using your expertise to shape sustainable transport and infrastructure solutions for the future. If you seek to make a difference on a global scale, working with next-gen technologies and the sharpest collaborative teams, then we could be a perfect match. 

Background of thesis project
At Volvo Trucks’ final assembly plants, a flexible internal logistics system based on autonomous mobile robots (AMRs) delivers materials just in time to the assembly line. These AMRs rely on ceiling-mounted cameras for perception, using object detection to identify obstacles and ensure collision-free navigation. Since multiple cameras often observe the same area from different viewpoints, a key challenge is how to fuse their inputs effectively to improve detection robustness and accuracy. A naïve approach is to run object detection independently on each camera view and fuse the results in post-processing. However, this strategy is highly sensitive to errors in individual views, which limits overall performance. In contrast, state-of-the-art methods perform end-to-end feature fusion, where features are extracted from each view, projected into a common representation, and processed jointly to generate detections. This is typically implemented as a deep neural network (DNN) trained across multiple calibrated cameras. While highly effective, such models rely on large labeled datasets and often fail to generalize to new deployments with different camera configurations.

To improve transferability, the thesis will explore Unsupervised Domain Adaptation (UDA) techniques to adapt trained models to new camera setups using unlabeled data. Specifically, the work will explore and develop training strategies and novel architectures for multi-view DNNs that enhance performance and adaptability, with the goal of making multi-view detection more suitable for real-world deployment. For academic rigor and comparability, experiments will be conducted on publicly available datasets for multi-view pedestrian detection. This project builds on MVUDA, an existing UDA baseline for multi-view pedestrian detection developed by industrial PhD student Erik Brorsson, who will serve as the project supervisor.

Suitable background
We are looking for MSc students with a strong interest in deep learning and computer vision. Candidates are expected to come from one of the following programmes (or similar):

• Systems, Control and Mechatronics
• Complex Adaptive Systems
• Data Science and AI
• Other related engineering or computer science disciplines

Required Skills: Proficiency in deep learning and Python programming
Meriting Experience: Experience in computer vision

Thesis project tasks

• Review state-of-the-art multi-view pedestrian detection methods with emphasis on deep neural network (DNN) architectures.
• Evaluate the robustness of selected architectures under domain shifts, such as sim-to-real discrepancies or changes in camera configuration.
• Develop improved architectures or training strategies to enhance adaptation across domains, for example through Unsupervised Domain Adaptation.
• (Optional) Summarize findings in a research paper with the aim of submission to a relevant academic venue.

Thesis Level
Master

Language
Thesis is to be written in English.

Starting date
Latest 19th of January 2026 

Number of students
Two (2)

Last day to apply: 2025-11-09

Tutor
Knut Åkesson, Chalmers,  knut.akesson@chalmers.se  (Supervisor)
Erik Brorsson, Volvo Group Trucks Operations erik.brorsson@volvo.com

Contact
Kristofer Bengtsson, Volvo Group Trucks Operations kristofer.bengtsson@volvo.com

References
AMR system at Volvo
https://www.youtube.com/watch?v=DA7lKiCdkCc

MVUDA 
https://github.com/ErikBrorsson/MVUDA
https://arxiv.org/abs/2412.04117

We value your data privacy and therefore do not accept applications via mail. 

Who we are and what we believe in 
We are committed to shaping the future landscape of efficient, safe, and sustainable transport solutions. Fulfilling our mission creates countless career opportunities for talents across the group’s leading brands and entities.

Applying to this job offers you the opportunity to join Volvo Group. Every day, you will be working with some of the sharpest and most creative brains in our field to be able to leave our society in better shape for the next generation. ​We are passionate about what we do, and we thrive on teamwork. ​We are almost 100,000 people united around the world by a culture of care, inclusiveness, and empowerment. 

Group Trucks Operations encompasses all production of the Group’s manufacturing of Volvo, Renault and Mack trucks, as well as engines and transmissions. We also orchestrate the spare parts distribution for Volvo Group’s customers globally and design, operate and optimize logistics and supply chains for all brands. We count 30,000 employees at 30 plants and 50 distribution centers across the globe. Our global footprint offers an opportunity for an international career in a state-of-the-art industrial environment, where continuous improvement is the foundation. As our planet is facing great challenges, we - one of the largest industrial organizations in the world - stand at the forefront of innovation. We are ready to rise to the challenge. Would you like to join us?