Structural Room Layout Point Clouds and Depth Maps

Point clouds and depth maps derived from the Hypersim dataset, re-rendered with all objects removed to retain only the first structural layout surface and its normal map. The dataset was created by user 'hugsam' and was last updated on March 6, 2026. It originates from Apple Inc.'s Hypersim dataset of photorealistic synthetic indoor scenes.

Use Cases

• Train monocular depth estimation models using the rendered depth maps of structural room envelopes.
• Develop 3D reconstruction algorithms from point clouds representing the first structural surface of a room.
• Benchmark scene understanding models on normal maps derived from room layout surfaces.
• Generate synthetic training data for tasks requiring object-agnostic room layout analysis.

Strengths

• Derived from the Hypersim dataset, which contains over 70,000 photorealistic images.
• Provides specialized data (point clouds, depth maps) focused purely on structural room envelopes.

Limitations

• Exact row count, file formats, and dataset size are unknown from the provided description.
• The dataset is synthetic, which may introduce a domain gap when applying models to real-world data.
• Limited to indoor scenes from the original Hypersim source.

Provenance

Source

Derived from Apple Inc.'s Hypersim dataset.

Collection Method

Created by processing Hypersim data to generate point clouds and re-render depth maps with objects removed.

Freshness

Last updated on March 6, 2026.