QoEVAVE Content Production: A High Quality 360° Video Dataset with Higher-Order Ambisonics Audio for Virtual Reality (online)

* Presenting author
Day / Time: 24.03.2022, 15:00-15:20
Room: 9-0208
Typ: Vortrag (strukturierte Sitzung)
Online-access: Bitte loggen Sie sich ein, damit weitere Inhalte sichtbar werden (bspw. der Zugang zur Onlinesitzung).

A new database of high-quality 360° audiovisual evaluation content is presented. Considerations towards capturing real-life-like audiovisual content are described, along with a review of practical recording and post-production challenges. The still-growing database currently consists of more than ten monoscopic videos captured in 8K (3840 x 7680) resolution with a statically installed Insta360 Pro 2 camera at 60 frames per second. The auditory counterpart is captured simultaneously with the em32 Eigenmike® spherical microphone array enabling up to 4th order Ambisonics. Audio and visual content is available independently at various frame rates, resolutions, or Ambisonic orders or as a single file in an mp4 container. The database includes various real-life scenes across nature and urban settings, capturing dynamic and static audio, visual, and audiovisual objects with at least 60-second average duration. Building on currently available mostly uni-modal databases, our content is a novel research contribution, targeting multi-modal aspects of human behavior, perception, and cognition. The scenes available to date are the first iteration of an ongoing production cycle and will also be further extended and augmented by interactive scenes with computer-animated video. The complete database shall enable a variety of scenarios for virtual reality experience research.