Master Thesis – Core Tech Graphics: Three Different Ideas

Master Thesis – Core Tech Graphics: Three Different Ideas

We have three ideas for you to pick from:

1. AI 3D Segmentation for Digital Twin

2. Visual-thermal automatic alignment

3. World Model for Depth Estimation and Segmentation in Body-Worn Camera

Category Computer Science, Applied Mathematics, Technical Physics, Electronics Engineering or similar

Scope Two Master students, 30hp

Idea 1: AI 3D Segmentation for Digital Twin

Dynamically updating a Digital Twin using monocular depth

This thesis explores the use of AI models to gain semantic understanding of the scene and depth information. This enables the possibility to keep track of changes to the scene that could be paired with metadata of objects.

To keep a Digital Twin updated, you will usually do a re-scan of the scene. A previous thesis project has shown that cameras can be used to update the scene instead of doing a re-scan. However, that project didn't really gain semantic understanding of the scene, which leads to a couple of limitations. By utilizing AI enabled 3D segmentation on a 2D image or image sequence, one can achieve more reliable upgrades to the Digital Twin models and keep it up to date.

Your goals

• Develop a reliable approach for detecting and segmenting scene changes.
• Compare modern monocular depth-estimation models and select one suited to the project.
• Measure how accurately new objects can be placed in a Digital Twin, in terms of depth and position.
• Explore how cameras can help keep Digital Twins up to date without repeated rescanning.

Idea 2: Automatic Alignment of Visual and Thermal Images

Bring visual detail and thermal information together

Some cameras capture both a visual image, rich in color and detail, and a thermal image that reveals heat. Combining the two views can provide more useful information, but differences in field of view and the way objects appear can make alignment challenging.

Alignment often depends on manually identifying a point shared by both images. This thesis will investigate whether computer vision can find corresponding features automatically and use them to align visual and thermal views.

Your goals

• Review research on automatic alignment of visual and thermal images.
• Implement or adapt methods for detecting shared features in paired images, then compare their performance.
• Identify the conditions that help or hinder alignment.
• Assess what data, calibration, or further development would be needed for a practical solution.
• Present a demonstration, findings, and recommendations for future work.

Idea 3: World Models for Depth Estimation and Segmentation in Body-Worn Camera Video

Towards more accurate 3D indoor reconstruction

Body-worn cameras record rooms, corridors, and objects as the wearer moves through a building. This footage could help create 3D indoor reconstructions, but camera movement, changing viewpoints, and occlusions make it difficult to understand the scene accurately.

V-JEPA is a video representation model developed in the context of world-model research. While it learns useful visual features from video, producing accurate depth maps and segmentation masks requires additional prediction modules and training. This thesis will explore how to adapt V-JEPA to improve those capabilities.

Your goals

• Improve depth estimation and segmentation in body-worn camera footage.
• Establish a baseline using pre-trained V-JEPA features and task-specific prediction modules.
• Test improvements to training objectives and ways of combining features from different model layers.
• Explore whether information across video frames can make predictions more consistent.
• Evaluate depth and segmentation results, then assess their effect on indoor reconstruction using an existing 3D pipeline. This cross-team project at Axis brings together practical programming and research in video world models and 3D computer vision. It’s an opportunity to apply emerging AI techniques to a real-world challenge.

Who to contact for any questions regarding the position:

Sara Garmark, hiring manager

Type of Employment

Thesis Worker (Fixed Term)

Posting End Date

2026-11-27

Certain roles at Axis require background checks, which means applicable verifications will be done in these recruitments. Notice will be provided before we take any action.

About Axis Communications

We enable a smarter and safer world by creating innovative solutions for improving security and business performance. As a network technology company and industry leader, we offer solutions in video surveillance, access control, intercom, and audio systems, enhanced by intelligent analytics applications.

With around 6,000 committed employees in over 50 countries, we collaborate with partners worldwide. Together, we thrive in our friendly, open, and collaborative culture and inspire each other to think beyond the expected. United by our commitment to inclusion, diversity, and sustainability, we consistently seek to develop our skills and way of working.

Let´s create a smarter and safer world

For more information about Axis, please visit our website www.axis.com.

Listen to Axis Communications Playlist