Multi-Camera Outdoor Scene Captures for Photometric Compensation

Four outdoor scenes were captured, each using three different cameras, for multi-view 3D reconstruction. The dataset was created by NVIDIA to accompany research on compensating for photometric variations in radiance field reconstruction. It includes photos processed with automatic exposure and color adjustments.

Use Cases

• Training models for novel view synthesis using multi-view images from three cameras per scene.
• Developing photometric compensation algorithms using exposure-bracketed images (+/-2 EV).
• Benchmarking 3D reconstruction methods on object-centric outdoor scene data with controlled camera variations.
• Researching automatic exposure and color processing effects on radiance field quality.

Strengths

• Contains four distinct outdoor scenes, providing multiple test cases.
• Each scene includes captures from three different cameras, enabling multi-camera analysis.
• Images feature exposure bracketing of +/-2 EV, offering photometric variation data.

Limitations

• Sample size is limited to only four scenes, which may restrict model generalization.
• The dataset is object-centric, potentially lacking broader environmental context.
• Specific row counts, image resolutions, and file formats are not provided in the description.

Provenance

Source

NVIDIA

Collection Method

Object-centric scene captures of outdoor scenes using multiple cameras with exposure bracketing.

Time Range

null

Freshness

null

Geography

Outdoor scenes (specific location unknown).