Vidhance is a Swedish software company based in Uppsala that develops video enhancement software for cameras in motion, such as smartphones, wearables and drones. 

Bellow is our 4 proposals for master thesis projects this year. If any sound interesting, get in contact with us!

Practical information for all proposals:

• Aimed at final-year master’s students in IT, Physics or a similar program 
• Runs during spring 2027 (January–June) 
• One or two students per thesis 
• Considerable freedom in how to approach the work 

Send your application to: career@vidhance.com
Subject Line: Master Thesis project  – [Your Name]
Include CV as an attached documents, not in the body of the email. If you want please also send grades and what project you are interested in.

Lightweight Patch Classification for UAV Threat Detection 

Vidhance has a detector that finds potential threats in the image. To be useful in a UAV defense setting, each detection also needs to be classified, and this must run on the limited compute hardware available on the platform. The goal of this thesis is to develop and evaluate a lightweight classifier that, as a downstream task, classifies the image patches extracted by the detector. The work includes: 

• Designing a framework for data collection and annotation. 
• Defining a classification output suited to the UAV defense use case. 
• Designing the classifier and evaluating its performance, including: 
• Classification metrics, such as per-class precision and recall 
• Profiling on UAV compute hardware 

GNSS-Free Navigation for UAVs/UGVs 

GNSS signals can be unavailable or unreliable, for example due to jamming or obstruction, so UAVs and UGVs need a way to navigate without them. The goal of this thesis is to design and evaluate a GNSS-free navigation feature for a UAV or UGV. The work includes: 

• Researching possible solutions, such as visual odometry or map matching, suitable for the limited compute hardware used in UAVs/UGVs. 
• Collecting an evaluation dataset with ground-truth positions. 
• Implementing the selected solution and evaluating it, including: 
• Navigation performance, such as position error and drift over time 
• Profiling on UAV/UGV compute hardware 

Gyro-Free Jello Effect Removal 

The jello (rolling shutter) effect degrades image quality, typically on UAV platforms. When a gyro is present and/or zoom levels are low, the jello effect is fairly easy to handle. However, some platforms lack accurate gyro data or run at high zoom, where gyro precision is insufficient. For these platforms, a gyro-based solution is not possible and a gyro-free solution is necessary. The goal of this thesis is to develop and evaluate a gyro-free jello effect filter. The work includes: 

• Collecting UAV video material, including footage from a gyro-equipped platform to use as a reference. 
• Researching and developing a gyro-free jello effect filter. 
• Evaluating the filter, including: 
• Image quality compared with a gyro-based reference 
• Profiling on UAV compute hardware 

Video Stitching with Relatively Rotating Cameras 

Platforms with more than one camera, for example a fixed camera and a gimbal-mounted camera, can provide a single combined view if their video streams are stitched together. The goal of this thesis is to research and design a feature that stitches video from two cameras that can rotate relative to each other. The algorithm needs to deliver good stitching quality on compute-limited platforms. The work includes: 

• Designing the hardware for the two-camera setup. 
• Researching, implementing and evaluating the feature. Evaluation includes: 
• Video stitching quality, such as alignment accuracy and visible seams 
• Profiling on UAV compute hardware