distributed and interactive systems logo

CWI DIS - Visual Saliency of Dynamic Point Clouds

Reference

Distorted

Front View

Back View

Introduction

This research introduces two complementary eye-tracking datasets for studying visual attention in Dynamic Point Clouds (DPC) within Virtual Reality environments. We compare task-free viewing (24 participants, 19 DPCs) with task-dependent quality assessment (40 participants, 5 reference DPCs with multiple distortions). Our analysis reveals significant differences in visual attention between these paradigms, measured using Pearson correlation and an adapted Earth Mover’s Distance metric. This work establishes a crucial connection between quality assessment and visual attention in DPCs, providing valuable benchmark data for improving VR experiences and visual saliency models.

Apparatus

Experimental Design

Data Processing

Data Collection: Angular Error Estimation

Gaze Point Identification

Truncated Cone-Sector

Each stage of our data processing pipeline is carefully designed to maintain data integrity while extracting meaningful insights about visual attention patterns in dynamic point cloud environments.

Comparison

Task-dependent

Results

Comparison Consistency of Visual Saliency Map Higher value means two saliency maps are more similar

Publications

  1. X. Zhou, I. Viola, E. Alexiou, J. Jansen, and P. Cesar QAVA-DPC: Eye-Tracking Based Quality Assessment and Visual Attention Dataset for Dynamic Point Cloud in 6 DoF. In IEEE International Symposium on Mixed and Augmented Reality (ISMAR), 2023.
    PDF
  2. X. Zhou, I. Viola, S. Rossi, and P. Cesar Comparison of Visual Saliency for Dynamic Point Clouds: Task-free vs. Task-dependent. IEEE Transactions on Visualization and Computer Graphics, 31(5): pp. 2964 - 2974, 2025. PDF
    Github Repository
    PDF
  3. X. Zhou, I. Viola, R. Yin, and P. Cesar Visual-Saliency Guided Multi-modal Learning for No Reference Point Cloud Quality Assessment. In Proceedings of the 3rd Workshop on Quality of Experience in Visual Multimedia Applications, 2024.
    PDF

Datasets

  1. Dynamic Point Cloud Quality Score & Eye-tracking Dataset (QAVQ-DPC) - Data from the 1st task-dependent user study
    LinkDataset 
  2. Dynamic Point Cloud Eye-tracking Dataset - Data from the 2nd task-free user study
    LinkDataset 

Contact

For questions about this research, please contact Xuemei Zhou (xuemei.zhou [at] cwi.nl)