360-degree Video Database for VR Sickness Assessment

Overview

For evaluation of the prediction performance of the objective VRSA, it is required to obtain subjective VR sickness scores such as SSQ and physical symptom scores (Oculomotor, Nausea, and Disorientation) for stimuli of VR contents. The prediction performance of the objective VRSA methods can be evaluated by measuring the correlation between the subjective VR sickness scores and the predicted scores obtained from objective VRSA. In this paper, we built a newly collected 360-degree video database, the corresponding subjective scores.

This web page provides the datasets of our IEEE TIP 2018 paper [1], so that researchers can repeat our experiments on our datasets. The dataset is for research purposes only. If you use our datasets, please cite the paper [1].

360-degree Video Dataset

In this paper, we collected a 360-degree video dataset with high spatial resolution and various motion patterns. In most of the previous works for subjectively and objectively assessing VR sickness, only one or two VR contents were used. We newly collected nine 360-degree videos including various scenes from Youtube as a benchmark and conducted an extensive subjective experiment for evaluation of objective VRSA. The videos contain various scenes such as beach, driving, flight, roller coaster, etc. To investigate the effect of various motion patterns on VR sickness, we collected 360-degree video datasets with various motion patterns from static to dynamic. They were subjectively divided into three categories based on motion patterns: simple, normal, and complex. The most of the collected contents have 4K resolution (3840 × 1920 or 3840 × 2048) with 30 Hz. Due to the viewing safety issue of the participated subjects, each test video was presented for 90 seconds

Table III (in our paper [1])

Simple 1 : [LINK] Simple 2 : [LINK] Simple 3 : [LINK]

Normal 4 : [LINK] Normal 5 : [LINK] Normal 6 : [LINK]

Complex 7 : [LINK] Complex 8 : [LINK] Complex 9 : [LINK]

Table VII (in our paper [1])

1 : [LINK] 2 : [LINK] 3 : [LINK] 4 : [LINK] 5 : [LINK] 6 : [LINK] 7 : [LINK] 8 : [LINK] 9 : [LINK] 10 : [LINK]

11 : [LINK] 12 : [LINK] 13 : [LINK] 14 : [LINK] 15 : [LINK] 16 : [LINK] 17 : [LINK] 18 : [LINK] 19 : [LINK] 20 : [LINK]

Subjective SSQ and Physical Symptom Scores

In subjective assessment experiments, Oculus Rift CV1 was used for displaying 360-degree videos, which was one of the high-end stereoscopic type HMDs. Its display resolution is 2160 × 1200 pixels (1080 × 1200 pixels per eye). Its display frame rate is maximum 90 Hz and it has 110 degree FoV. A total of twenty subjects, aged 20 to 30, participated in our subjective experiments under the approval of KAIST Institutional Review Board (IRB). In general, the use of VR is not recommended for young people under 12 years of age due to immature development of visual-vestibular sensors. The participants in our experiment do not have health problems such as immature development of visual-vestibular sensors, vestibular dysfunction or oculomotor dysfunction, compared to children and older people. Subjects have normal or corrected-to-normal vision and minimum stereopsis of 60 arcsec. In our experiment, before watching each stimulus, they were placed in the center position to be started from zero position in order to prevent significantly different viewing traces between viewers They were seated on a rotatable chair in order to freely look around 360-degree contents. A week before the actual subjective assessment experiments, we had subjects experience a variety of VR contents with Oculus Rift in order to allow them to familiarize with VR environment. In our experiments, the subject head motion was small and negligible during watching 360-degree contents. Since most of the 360 degree-videos used in our experiment have movement in a certain direction by roller coaster and car, subjects focused their gaze in the similar direction (e.g., the direction of rails in the roller coaster video or moving direction in the driving video). The head motion below the range of 44º to 55º in yaw could not cause severe VR sickness. All experimental environments followed the guideline as per the recommendations of ITU-R BT.500-13 and BT.2021.

Subjective SSQ and physical symptom (Oculomotor, Nausea, and Disorientation) score datasets : [Link]

Reference

If you use the database, please cite as :

[1] Kim, H. G., Lim, H. T., Lee, S., & Ro, Y. M. (2018). Vrsa net: Vr sickness assessment considering exceptional motion for 360 vr video. IEEE transactions on image processing, 28(4), 1646-1660.

[2] Kim, H. G., Lee, S., Kim, S., Lim, H.- taek, & Ro, Y. M. (2021). Towards a Better Understanding of VR Sickness: Physical Symptom Prediction for VR Contents. Proceedings of the AAAI Conference on Artificial Intelligence, 35(1), 836-844.