Enhancing Dynamic Point Clouds in the Wild: A Grand Challenge on Real-World 4D Volumetric Data

Xuemei Zhou, Jack Jansen, Pablo Cesar, Guillaume Gautier, Alexandre Mercat, Jarno Vanne, Irene Viola

Keywords: point cloud enhancement, real world capture, dynamic point cloud

TL;DR: This Grand Challenge introduces an in-the-wild benchmark for dynamic colored point cloud enhancement with paired low-quality captures and high-quality ground truth, targeting denoising, completion, upsampling, and temporal consistency.

Abstract:

Dynamic point clouds captured in real-world environments are fun-
damental to immersive multimedia applications such as volumetric
video, XR telepresence, and digital twins. However, point clouds ac-
quired by consumer-grade sensors suffer from severe degradations,
including noise, sparsity, missing geometry, temporal instability, and
color artifacts, which significantly limit downstream reconstruction,
rendering, and compression. Existing point cloud enhancement meth-
ods are predominantly evaluated on synthetic benchmarks and static
scenes, leaving a critical gap in systematic evaluation for real-world,
dynamic (4D), and color point clouds. This Grand Challenge intro-
duces the first in-the-wild benchmark for dynamic point cloud en-
hancement based on the UVG-CWI-DQPC dataset, which provides
paired low-quality consumer-grade captures and high-fidelity multi-
sensor ground truth across diverse dynamic human-centric sequences.
The challenge targets unified enhancement of denoising, completion,
and upsampling, while explicitly accounting for temporal consistency
and color fidelity. Participants are evaluated using a comprehensive
protocol combining geometric accuracy, perceptual quality, tempo-
ral stability, and computational efficiency, with the top submissions
further assessed via controlled subjective studies. This challenge
aims to foster realistic algorithm design, fair comparison, and accel-
erated progress toward practical deployment of dynamic point cloud
enhancement in multimedia systems.

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Submission Number: 5