Master Thesis – Enhanced Reasoning in Vision-Language Models Using Structured Visual Representations

Vision-Language Models often reason in language after processing images, which can make detailed visual information harder to retain and use. Surveillance systems also produce structured data, including object identities, bounding boxes, tracks and velocity vectors. This thesis at Axis Communications will explore how such information can help VLMs interpret video and answer questions about objects, spatial relationships and changes over time.

The assignment

• Review current research on integrating structured visual information into VLMs.
• Explore surveillance metadata and ways to generate training data from internal and public datasets.
• Fine-tune one or more open-source VLMs through approaches such as instruction fine-tuning and parameter-efficient methods, including LoRA.
• Investigate how structured information can be represented, aligned with visual inputs and used during training, inference and intermediate reasoning.
• Evaluate performance on questions involving objects, spatial relationships and temporal events, comparing models trained with and without structured visual information.
• Assess the interpretability and limitations of the proposed approach.

Who we are looking for

• Master’s students in Electronics Engineering, Mathematics, Computer Science, Biomedical Engineering or Engineering Physics.
• Strong interest in machine learning, Vision-Language Models, programming and problem-solving.
• Affiliated with a Swedish university or university college, either directly or through an exchange programme.

What we offer

• An open, friendly and collaborative company culture that encourages new ideas and continued development.

Practical details

Location: Lund. The project is intended for two master’s students and covers 30 credits per student. Apply through the thesis advert by 27 November 2026; applications may be written in Swedish or English. Apply as a pair, submitting separate applications and naming your co-applicant. Include a CV, cover letter and university grade transcript. Questions: Magnus Svensson, Engineering Manager, Magnus.Svensson@axis.com, +46 46 2879187.