SynMVCrowd: A Large Synthetic Benchmark for Multi-view Crowd Counting and Localization

2026-03-25

AI Summary

Existing methods are predominantly evaluated on small-scale, limited-viewpoint, and sparsely populated scenes, which fail to reflect performance in real-world complex environments. To address this gap, this work introduces SynMVCrowd, the first synthetic benchmark tailored for large-scale multi-view crowd analysis, comprising 50 high-fidelity 3D scenes with multi-view imagery and densely packed crowds of up to 1,000 individuals. Leveraging this benchmark, we develop a strong baseline model that effectively fuses multi-view information for accurate localization. Experiments demonstrate that our approach not only outperforms existing methods on SynMVCrowd but also significantly enhances domain transferability to novel real-world scenes and improves single-image counting accuracy.

Technology Category

• Computer Vision: Scene Analysis & Understanding
• Machine Learning: Multi-instance/Multi-view Learning
• Humans and AI: Crowd Sourcing and Human Computation

Application Category

• Security and Privacy: Large-scale security measurements
• User Modeling, Personalization and Recommendation: Practical large-scale studies of user experience
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Existing multi-view crowd counting and localization methods are evaluated under relatively small scenes with limited crowd numbers, camera views, and frames. This makes the evaluation and comparison of existing methods impractical, as small datasets are easily overfit by these methods. To avoid these issues, 3DROM proposes a data augmentation method. Instead, in this paper, we propose a large synthetic benchmark, SynMVCrowd, for more practical evaluation and comparison of multi-view crowd counting and localization tasks. The SynMVCrowd benchmark consists of 50 synthetic scenes with a large number of multi-view frames and camera views and a much larger crowd number (up to 1000), which is more suitable for large-scene multi-view crowd vision tasks. Besides, we propose strong multi-view crowd localization and counting baselines that outperform all comparison methods on the new SynMVCrowd benchmark. Moreover, we prove that better domain transferring multi-view and single-image counting performance could be achieved with the aid of the benchmark on novel new real scenes. As a result, the proposed benchmark could advance the research for multi-view and single-image crowd counting and localization to more practical applications. The codes and datasets are here: https://github.com/zqyq/SynMVCrowd.

Problem

• multi-view crowd counting
• crowd localization
• synthetic benchmark
• large-scale scene
• dataset limitation

Innovation

• synthetic benchmark
• multi-view crowd counting
• crowd localization
• domain transfer
• large-scale dataset