ToF-360: Spherical RGB-D Images with Instance and Layout Annotations

179 equirectangular RGB images with corresponding depth, surface normals, XYZ, and HHA images. The dataset, created by COLE-Ricoh, provides instance-level semantic and room layout annotations for 4 unique scenes. It was last updated on Hugging Face in June 2025.

Use Cases

• Train instance segmentation models based on the provided building-defining object category labels.
• Develop room layout estimation algorithms based on the annotated ceiling-wall and wall-floor boundaries.
• Research multimodal scene understanding based on the aligned RGB, depth, surface normal, and HHA image modalities.
• Benchmark 360-degree perception systems based on the spherical equirectangular image format.

Strengths

• Contains 179 high-resolution equirectangular RGB images.
• Provides aligned depth, surface normal, XYZ, and HHA image data for each RGB frame.
• Includes instance-level semantic annotations and room layout boundaries for 4 distinct scenes.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and dataset size are unknown, which may limit suitability assessment.

Provenance

Source

COLE-Ricoh

Freshness

Last updated 2025-06-06 22:20:00