Blender NeRF Dataset Pipeline

Blender scripts for NeRF dataset generation.

Camera Trajectory Mode

• random_hemisphere : uniformly sample the surface of a hemisphere looking at the object.
• random_sphere : uniformly sample the surface of a sphere looking at the object.
• circular : evenly sampling a circular path around the object.
• figure_eight: evenly sampling a figure 8 path with object at the center.

Usage

Tested with Blender 3.3

object_nerf.py: Creates dataset with only the .obj model. Model is illuminated by white background lighting.

composite_nerf.py: Creates dataset with the .obj model sitting on a plane. Scene is illuminated by the HDRI specified.

config.blend: Scripts will use the render configurations in this empty blender file. Settings you might want to change:

• Rendering Device and backend
• Output image resolution. Default 800 x 800
• Camera focal length
• Render setting such as SPP, Max Bounces, Depth of Field.

Example: Object

nerf-pipeline> blender --background config.blend --python scripts\object_nerf.py -- --object models\bench3\model.obj --views 50 --mode random_sphere --output dataset\bench3

Depending on your setup you might have to enter the full path to the Blender executable.

Example: Composite

nerf-pipeline> blender --background config.blend --python scripts\composite_nerf.py -- --object models\bench1\model.obj --views 60 --mode random_hemisphere --output dataset\bench_hdri --hdri hdri\daylight.exr