leaf-mechanics
收藏资源简介:
该数据集源自2022年发表于《自然》合作期刊的研究《Nature-inspired architected materials using unsupervised deep learning》(作者:S. Shen, M. Buehler),旨在支持受自然启发的架构材料研究,特别是与无监督深度学习相关的任务。数据内容包含材料微结构或设计图像(‘image’字段)及其关联的物理属性数据,具体特征包括:图像数据(‘image’)、文件名(‘filename’)、密度(‘density’,浮点型)和归一化模量(‘mod_norm’,浮点型)。数据集被划分为训练集和测试集,其中训练集包含4000个样本,测试集包含1000个样本。适用于材料科学、计算材料学和机器学习交叉领域的研究,例如材料性能预测、微结构-性能关系建模或基于深度学习的材料生成与设计。
This dataset originates from a 2022 study published in a Nature partner journal titled Nature-inspired architected materials using unsupervised deep learning (authors: S. Shen, M. Buehler). It aims to support research on nature-inspired architected materials, particularly tasks related to unsupervised deep learning. The data includes material microstructure or design images (the image field) and associated physical property data, with specific features such as image data (image), filename (filename), density (density, floating-point type), and normalized modulus (mod_norm, floating-point type). The dataset is split into training and test sets, with the training set containing 4000 samples and the test set containing 1000 samples. It is suitable for research in interdisciplinary fields of materials science, computational materials science, and machine learning, such as material property prediction, microstructure-property relationship modeling, or deep learning-based material generation and design.
数据集:leaf-mechanics
来源:MIT Laboratory for Atomistic and Molecular Mechanics (LAMM)
描述:该数据集包含叶状结构材料的图像及其对应的力学性能参数,用于基于无监督深度学习的仿生材料研究。
数据特征:
image:图像数据(image 类型)filename:文件名(string 类型)density:密度(float64 类型)mod_norm:归一化模量(float64 类型)
数据划分:
- 训练集:4000 个样本,大小约 49.95 MB
- 测试集:1000 个样本,大小约 12.77 MB
数据集总大小:约 62.74 MB(下载大小约 62.54 MB)
引用来源:
S. Shen, M. Buehler, "Nature-inspired architected materials using unsupervised deep learning", Communications Engineering, 2022.
论文链接:https://www.nature.com/articles/s44172-022-00037-0
配置:
- 默认配置(
default),训练数据路径为data/train-*,测试数据路径为data/test-*。




