ShapeSplats Sharing: 2D Image, Depth, and Normal Renders of ShapeNet Objects

ShapeSplats provides a collection of 2D image, depth, and normal map renders derived from the ShapeNet 3D object repository. The dataset includes per-view camera parameters stored in a structured JSON format for each object. It was last updated on September 1, 2025.

Use Cases

• Training novel view synthesis models based on the provided multi-view image renders.
• Developing depth estimation algorithms using the included per-view depth maps.
• Benchmarking 3D reconstruction methods based on the paired image and normal map data.
• Calibrating or testing camera pose estimation techniques using the detailed camera parameter JSON files.

Strengths

• Includes multiple synchronized render types (image, depth, normal) per view, enabling multi-task learning.
• Camera parameters for each view are provided in a detailed JSON format, offering precise scene geometry.

Limitations

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

Provenance

Source

ShapeSplats

Collection Method

Rendered from the ShapeNet 3D object dataset.

Freshness

Last updated 2025-09-01 18:20:51; freshness should be verified.