ViVo

Introduction

We are in the process of formating data for distribution - for now we have uploaded temporary links to the 3D reconstruction and Compression tests (working on it...)

Hello! Below you will find samples for the scenes that we offer. We also illustrate the 14 multi-view video angles for each capture session and present examples of the visual data-types that are captured/generated using our dataset and code.

- If you are looking for the code documentation, see: Docs
- If you are looking for the paper results, see: Paper
- If you want to browse the scenes we captured, see: Catalogue

Scene Samples (from Paper)

Core Features

The dataset mainly captures music, dance and sport. There are also some special scenes that cover other forms of "unique" entertainment. Consequently, the dataset contains a variety of features and inter-mixes a variety of these within each scene. This includes but is not limited to:
- Highly reflective surfaces (cymbals, triangle, ...)
- Transparent and semi-transparent objects (glasses, key-chains, ...)
- Dynamic textures (fire, liquid, ...)
- Non-rigid deformations (e.g. balloons deflating)

Multi-person
Music
Multi-person
Dance and Sport
Multi-person
Specials
Multi-person
Multi-Person
Multi-person
Transparent, Reflective & Dynamic Textures
Multi-person
Animal

Multi-View Cameras

The data consists of 14x RGB and Depth video-camera pairs. The scenes were captures over a total of three sessions with varying lighting conditions. In the examples below we show the multi-view camera placement for each session. Note: the lighting within each session is fixed.

Session 1
Session 1
Session 2
Session 2
Session 3
Session 3

Visual Data

The raw data and data generated via the handling scripts, includes:
- RGB and Depth videos with Per-frame Intrinsics and Extrinsics (raw)
- Per-frame point clouds (generated)
- 2-D Masks (generated)

Point Clouds

Contributors

Contributor 1
Adrian Azzarelli
Contributor 2
Ge Gao
Contributor 3
Ho Man Kwan
Contributor 4
Fan (Aaron) Zhang
Contributor 5
Nantheera Anantrasirichai
Contributor 6
Dave Bull
Contributor 7
Olliver Moolan-Feroze

Cite Us

@article{azzarelli2025vivo,
    title={ViVo: A Dataset for Volumetric VideoReconstruction and Compression},
    author={Azzarelli, Adrian and Gao, Ge and Kwan, Ho Man and Zhang, Fan and Anantrasirichai, Nantheera and Moolan-Feroze, Ollie and Bull, David},
    journal={arXiv preprint arXiv:2506.00558},
    year={2025}
}