Uppmärksamhetsanalys i en Mixed Reality-miljö

This thesis explores how user attention in mixed reality can be analysed by combining eye tracking with object detection. A real-time module was developed for Microsoft HoloLens 2 to capture gaze data, project fixation points onto videoframes, and check for overlaps with detected objects. In this context, “attention” is defined as the user’s visual focus point. The module streams video and gaze data to a YOLOv5 object detection server. Due to hardware constraints, the server processes around 3 frames per second, which is lower than ideal for real-time applications. A user study was conducted in a controlled environment containing both virtual and physical objects. No explicit instructions were given about task elements to avoid influencing natural attention patterns.The module detects object-gaze overlaps and records metrics like fixation duration and object distance. While performance is currently limited by processing speed, the module demonstrates a working prototype for analysing gaze behaviour in mixed reality environments.