"Guangzhou Street"

Read full text Summarise with VIRAT
Authors: Bifa Liang
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

"The Guangzhou Street Dataset is a real-world binocular stereo vision dataset specifically curated to advance research in stereo matching and dense depth estimation. Captured within the complex and dynamic urban environments of Guangzhou, China, this locally sourced dataset provides high-quality stereo image pairs depicting a wide array of typical urban infrastructure. Key visual elements include vehicles, pedestrians, traffic poles, and street-level architectural structures, which inherently encompass challenging matching scenarios such as occlusions, textureless surfaces, and sharp object boundaries.The primary objective of this dataset is to serve as a robust evaluation benchmark for verifying the cross-domain adaptability and generalization capabilities of deep learning-based stereo networks in authentic, real-world deployments. By bridging the gap between standard synthetic or idealized training benchmarks and highly variable physical environments, the Guangzhou Street Dataset enables researchers to rigorously validate the reliability of intelligent vision systems. It is particularly valuable for evaluating depth estimation models tailored for autonomous driving, smart city infrastructure, and resource-constrained edge computing applications."

Fields of Study & Subjects

Advanced Vision and Imaging Advanced Neural Network Applications Autonomous Vehicle Technology and Safety

Publication details

Journal / source
IEEE DataPort
Publication date
21 March 2026
Language
EN
Document type
Dataset