Cloud4D: High-Resolution Stereo Imagery for 3D Cloud Reconstruction

Cloud4D contains high-resolution stereo camera captures of clouds for 3D reconstruction and depth estimation, released for a NeurIPS 2025 Spotlight paper. The dataset features synchronized multi-perspective imagery organized by date and hour to support high-resolution atmospheric modeling.

Use Cases

• Depth estimation using synchronized left_images and right_images pairs
• 3D cloud reconstruction from multi-view data found in perspective folders
• Temporal cloud motion analysis using the YYYYMMDD_HH directory sequences

Strengths

• NeurIPS 2025 Spotlight peer-reviewed quality
• Multi-perspective stereo camera captures for 3D modeling
• High temporal resolution with hourly folder organization

Limitations

• Geographic scope is limited to the United States
• Restricted to visible-light atmospheric conditions
• Total record count and dataset size are not explicitly stated

Provenance

Source

NeurIPS 2025 (Spotlight) publication by Jacob Lin et al.

Collection Method

Ground-based stereo camera sensors

Time Range

2023

Freshness

Last updated January 2026; contains data from at least 2023.

Geography

United States