VOLVQAD is a volumetric video quality assessment dataset consisting 7,680 ratings from 120 users on 376 video sequences. The volumetric video sequences are encoded with MPEG V-PCC using 4 different avatar models and 16 varying quality. The volumetric video sequences are then rendered into test videos for quality assessment using 2 different background colors and 16 different quality switching patterns.
The user ratings can be found in the file ratings.csv in this repo.
The file contains six columns:
- User ID: P001 to P120.
- Point Cloud Model: loot, redandblack, soldier, and longdress
- Quality Switching Pattern: ConstantR$x$, Cascade, InverseCascade, SpikeR$x$, and InverseSpikeR$x$
- Quality Feature: geometry (for varying geometry quality only), texture (for varying texture quality only), or both (for varying both geometry and texture quality)
- Background Color: black or gray
- Rating: User rating of this video
A paper on this dataset is published in ACM MMSys'23. If you use the ratings data from our paper, please cite
@InProceedings{
title = "VOLVQAD: An MPEG V-PCC Volumetric Video Quality Assessment Dataset",
author = "Samuel Rhys Cox and May Lim and Wei Tsang Ooi",
booktitle = "Proceedings of the 14th ACM International Conference of Multimedia Systems (MMSys)",
year = 2023,
month = June,
address = "Vancouver, Canada",
}
The 376 test video sequences are derived from the 8i Voxelized Full Bodies (8iVFB v2) Point Cloud Dataset (license)
Eugene d'Eon, Bob Harrison, Taos Myers, and Philip A. Chou, "8i Voxelized Full Bodies - A Voxelized Point Cloud Dataset," ISO/IEC JTC1/SC29 Joint WG11/WG1 (MPEG/JPEG) input document WG11M40059/WG1M74006, Geneva, January 2017.
The test sequence is about 1.5 GB. Request to access the test video sequence can be made to [email protected]