RGBN Datasets



The following are datasets of "RGBN" - images with a color and a normal per pixel. Many of these first appeared in the following paper:

Corey Toler-Franklin, Adam Finkelstein, and Szymon Rusinkiewicz.
"Illustration of Complex Real-World Objects using Images with Normals."
International Symposium on Non-Photorealistic Animation and Rendering (NPAR), August 2007.

These datasets may be used, with acknowledgment, for noncommercial purposes only.
   − Szymon Rusinkiewicz, Corey Toler-Franklin, and Adam Finkelstein


Dataset: Leaf of rainbow chard

Download: chard.zip

Dataset: Leaf of rainbow chard #2

Download: chard2.zip

Dataset: Cutting board with vegetables

Download: food.zip

Dataset: Rhododendron leaves and bud

Download: leaves.zip

Dataset: Pine cone

Download: pinecone3.zip

Dataset: Conch shell

Download: shell.zip

Dataset: Terra cotta soldier miniature

Download: soldier.zip

Dataset: Tools

Download: tools.zip

Dataset: Tools #2

Download: tools2.zip

Dataset: Log, pine cones, and conifer needles

Download: wood.zip


Technical information: These datasets were acquired with photometric stereo. The original images were captured using a Canon 5D camera and hand-held flash. The data are distributed as PPM files encoding normals (the RGB channels hold the X, Y, and Z components of the normal - the range [0..255] maps linearly to [-1..1]) and colors (gamma 2.2). For most datasets, a discontinuity map is also included.