The Mokume Dataset

This dataset is associated with the journal paper "The Mokume Dataset and Inverse Modeling of Solid Wood Textures" (PDF, SIGGRAPH 2025). The data includes the Mokume Dataset, which captures real data of wood textures. It also includes a trained model (U-Net) for extracting annual rings from photographs.

Abstract

We present the Mokume dataset for solid wood texturing consisting of 190 cube-shaped samples of various hard and softwood species documented by high-resolution exterior photographs, annual ring annotations, and volumetric computed tomography (CT) scans. A subset of samples further includes photographs along slanted cuts through the cube for validation purposes.

Using this dataset, we propose a three-stage inverse modeling pipeline to infer solid wood textures using only exterior photographs. Our method begins by evaluating a neural model to localize year rings on the cube face photographs. We then extend these exterior 2D observations into a globally consistent 3D representation by optimizing a procedural growth field using a novel iso-contour loss. Finally, we synthesize a detailed volumetric color texture from the growth field. For this last step, we propose two methods with different efficiency and quality characteristics: a fast inverse procedural texture method, and a neural cellular automaton (NCA). 

We demonstrate the synergy between the Mokume dataset and the proposed algorithms through comprehensive comparisons with unseen captured data. We also present experiments demonstrating the efficiency of our pipeline’s components against ablations and baselines. 

Folder Structure & Contents

There are three items: 

• MokumeDataset (approx. 20,000 MB). This folder contains the raw data, including high-resolution photographs, annotations, and more, of the documented 190 cube samples from 17 different species.
• ImageParis (approx. 240 MB). This folder contrains image pairs formated for the image translation task of converting a wood color image into an annual ring localization image.
• The file unet_trained_model.pt contains a trained U-Net, which takes a 64x64 pixel patch of a color photograph of wood and translates it to an annual ring licalization image. It is trained on the image pairs in the above folder.

Refer to sections below for detailed information about the contents of the folders.

MokumeDataset Folder

Inside the MokumeDataset folder, there are 190 subfolders, one for each sample, named XXNN where 

• XX = species code of 1 or 2 characters (e.g., "B", see "Wood Species Codes" below).
• NN = samples number (e.g., "02")

File content of each subfolder XXNN:

For more information, refer to the paper (PDF), in particular:

• The cube unfolding convention Fig. 29.
• The annual ring annotation image structure in Fig. 30.
• Details about the slanted cut surface position in the cube in Fig. 31. 

ImagePairs Folder

The ImagePairData folder contains formated data for the image translation task of identifying annual ring patterns in a color photo of wood. Inside this folder, there are two subfolders: test_data and training_data. Each folder contains image pairs:

where * is a unique image identifier. This data is created based on the raw data in the MokumeDataset folder. In particular, the color images are resized to smaller and exactly square images, and annual ring annotations are processed into annual ring localization images, which is a smooth representation of the annual rings, suitable for optimization/learning. For more information about how the localization images were created, refer to the paper (PDF), in particular Section 5.1.1 Training data and Fig. 12.

Wood Species Codes

Other info

Project page

For more information, visit: https://mokumeproject.github.io/ 

Reference

Maria Larsson, Hodaka Yamaguchi, Ehsan Pajouheshgar, I-Chao Shen, Kenji Tojo, Chia-Ming Chang, Lars Hansson, Olof Broman, Takashi Ijiri, Ariel Shamir, Wenzel Jakob, and Takeo Igarashi. 2025. The Mokume Dataset and Inverse Modeling of Solid Wood Textures. ACM Trans. Graph. 44, 4 (August 2025), 18 pages. https://doi.org/10.1145/3730874