metadata
dataset_info:
features:
- name: image
dtype: image
- name: label
dtype:
class_label:
names:
'0': AmbientOcclusion
'1': Color
'2': Displacement
'3': Metalness
'4': NormalDX
'5': NormalGL
'6': Opacity
'7': Roughness
splits:
- name: train
num_bytes: 1843104179
num_examples: 843
- name: validation
num_bytes: 484390783
num_examples: 211
download_size: 2327625226
dataset_size: 2327494962
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: validation
path: data/validation-*
license: cc-by-4.0
task_categories:
- image-classification
- zero-shot-image-classification
- image-to-text
language:
- en
tags:
- blender
- textures
- materials
pretty_name: Textures
size_categories:
- 1K<n<10K
Description
This is the third iteration and official release of a dataset curated to power the Materializer model for Blender. The dataset contains a range of labeled texture images that were sourced from ambientCG under their Creative Commons CC0 1.0 Universal License. These textures are designed to help in the classification of various material maps, which are essential for creating realistic 3D materials in Blender.
Future Plans
The dataset is still evolving, and I plan to expand it with a wider range of textures sourced from various online repositories. Feedback and contributions are welcome to help improve the dataset further.
References & Attributions
Created using various 1k images from ambientCG.com, licensed under the Creative Commons CC0 1.0 Universal License.