360-Degree Video Head Movement Dataset

Abstract

International audienceWhile Virtual Reality applications are increasingly attracting the attention of developers and business analysts, the behaviour of users watching 360-degree (i.e. omnidirectional) videos has not been thoroughly studied yet. This paper introduces a dataset of head movements of users watching 360-degree videos on a Head-Mounted Display (HMD). The dataset includes data collected from 59 users watching five 70 s-long 360-degree videos on the Razer OSVR HDK2 HMD. The selected videos span a wide range of 360-degree content for which different viewer's involvement, thus navigation patterns, could be expected. We describe the open-source software developed to produce the dataset and present the test material and viewing conditions considered during the data acquisition. Finally, we show some examples of statistics that can be extracted from the collected data, for a content-dependent analysis of users' navigation patterns. The source code of the software used to collect the data has been made publicly available, together with the entire dataset, to enable the community to extend the dataset

Visible Light Communication for Next Generation Untethered Virtual Reality Systems

“…A user is assigned a transmitter from which it has the shortest distance. Existing systems like [36]. We observed that the pitch and roll values stay within -55 o to 55 o in 96.13% and 99.43% of the cases respectively.…”

Deep Learning for Content-Based Personalized Viewport Prediction of 360-Degree VR Videos

“…The position-related data includes the VR user's head orientation, while the content-related data is each VR video frame itself. Simulation results using real data show that the proposed CNN-LSTM architecture can achieve high prediction accuracy, in particular, the proposed content-based approach can yield up to 16.1% compared to a baseline that relies only on position.…”

“…There exist a few datasets of head orientation for VR users. We explore one public head movement dataset for 360°videos [16]. This dataset includes 5 videos viewed by 59 users.…”

A taxonomy and dataset for 360° videos

“…Several datasets provide head movement traces of users watching 360 • videos. Corbillon et al [2] gathered a dataset of viewport traces for ve videos with 59 participants. Lo et al provided a dataset captured with 50 subjects watching 10 videos [11].…”