CHRNet Guangzhou Street Stereo Dataset

This dataset contains rectified, synchronized binocular RGB stereo pairs captured across diverse urban streets in Guangzhou, covering vehicles, buildings, fences, and sign poles. Each capture includes camera calibration and metadata (capture location and timestamps), enabling evaluation and fine-tuning of stereo depth and disparity estimation methods on real-world urban scenes. The data were used to validate CHRNet\u2019s outdoor performance and qualitative results reported in the manuscript.