PanoCity: Aerial Panoramic Images and Depth Maps from Three Chinese Cities

Beijing, Jinan, and Ningbo are covered by this 1.4 TB dataset containing 119,537 panoramic RGB images and an equal number of corresponding depth maps, each at 2048×4096 resolution. The dataset was created by YijingGuo and is hosted on Hugging Face. It was last updated on April 19, 2026.

Use Cases

• Train depth estimation models based on the paired panoramic RGB and depth map data.
• Develop 3D scene reconstruction algorithms for urban environments using the aerial panoramic imagery.
• Benchmark semantic segmentation or object detection models on large-scale, high-resolution panoramic scenes.
• Conduct geographic or urban planning analysis using the geospatial data from three distinct Chinese cities.

Strengths

• Large scale with 239,074 total image files (RGB + depth) totaling 1.4 TB.
• High-resolution panoramic images and depth maps at 2048×4096 pixels.
• Geographic diversity across 137 city blocks in Beijing, Jinan, and Ningbo.
• Includes official train/test splits as indicated by a splits_config.json file.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

huggingface

Freshness

Last updated 2026-04-19 18:26:20; freshness should be verified.

Geography

Beijing (20 blocks), Jinan (76 blocks), and Ningbo (41 blocks), China.