Anisotropic Spheres (ASPH)

Overview

The Anisotropic Spheres (ASPH) dataset is a synthetic benchmark dataset designed for evaluating neural rendering methods and inverse rendering algorithms that handle anisotropic reflectance under complex illumination. This dataset was created to support the research presented in ShinyNeRF.

Dataset Description

ASPH contains controlled renderings of four distinct sphere instances, each rendered under multiple high-dynamic-range (HDR) environment maps. The dataset varies environment maps to study their effects on appearance under different lighting conditions.

Key Features

• Simple, controlled geometry: Sphere primitives eliminate geometric complexity, allowing researchers to focus exclusively on material and illumination interactions
• Systematic parameter variations: Structured sweeps across anisotropic and roughness material properties enable controlled experiments on each object
• Multiple illumination conditions: Each sphere is rendered under diverse HDR environment maps to capture behavior across different lighting scenarios
• High-quality ground truth: Physically-based rendering provides accurate reference data for benchmarking

Dataset Contents

• RGB renderings: Rendered images (270×270 pixels) for each sphere-environment combination
• Material property maps: Ground truth maps including RGB, surface normals, tangent vectors, anisotropy (ani), roughness (ro), and depth (alpha)
• Blender scene files: Complete scene setups for reproducibility and custom rendering configurations