Quality of Multimedia Experience Meets Machine Intelligence

The “User-centric Analysis of Multimedia Data” Group presented its research at the International Conference on Quality of Multimedia Experience (QoMEX 2026)

360° videos are increasingly becoming a powerful medium for remote inspection and remote control. However, omnidirectional datasets for training and evaluating human and algorithmic performance remain rare, as their creation involves significant qualitative and technical challenges.

As part of the demo track, Florian Schlechte presented 360-FORGE—a CGI engine for omnidirectional ground-truth data in 360° annotation studies. The engine combines two key approaches:

1. A configurable Python framework: This controls the generation and movement paths within the generated CGI scenes. This delivers precise ground-truth data for each object, which is currently not feasible with real 360° footage.
2. A methodology optimised for crowdsourcing: implemented as a specialised campaign template, it offers a browser-based annotation interface that runs across platforms on laptops, haptic controllers and mobile devices.

Through intensive dialogue with subject matter experts from related disciplines, Florian Schlechte was able to gather valuable new insights for his future research.