Outdoor Scene Captures For Photometric Compensation In Radiance Fields

PPISP dataset accompanies the NVIDIA research paper 'PPISP: Physically-Plausible Compensation and Control of Photometric Variations in Radiance Field Reconstruction'. It contains object-centric captures of four outdoor scenes, each recorded with three different cameras. The dataset was uploaded to Hugging Face by NVIDIA in March 2026.

Use Cases

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

Strengths

• Contains captures of four distinct outdoor scenes.
• Each scene includes data from three different camera models.
• Images feature exposure bracketing of +/-2 EV for photometric analysis.

Limitations

• The exact number of images, rows, or total data size is unspecified.
• Limited to only four outdoor scenes, which may restrict model generalization testing.
• Geographic and environmental diversity of the scenes is not detailed.

Provenance

Source

NVIDIA, associated with the 'PhysicalAI-NuRec-PPISP' research project.

Collection Method

Object-centric scene captures using physical cameras with exposure bracketing, followed by reprocessing.

Freshness

Last updated on the platform in March 2026.

Geography

Outdoor scenes, specific locations unknown.