Molecular-MNIST

Molecular-MNIST数据集由国家能源研究科学计算中心创建，包含10种不同大小的分子，每个分子有2000种结构变体，总计20000条数据。数据集通过分子动力学轨迹的2D截面投影生成，包含形状图像和衍射图案。创建过程中，3D原子分布被投影到2D平面，并计算了傅里叶变换以生成衍射图案。该数据集主要用于机器学习和深度学习在散射、成像和显微镜研究中的应用，旨在解决快速实时推理的问题。

The Molecular-MNIST dataset was created by the National Energy Research Scientific Computing Center. It contains 10 distinct molecular sizes, each with 2000 structural variants, totaling 20,000 data entries. The dataset is generated via 2D cross-sectional projections of molecular dynamics trajectories, and includes both shape images and diffraction patterns. During the creation process, 3D atomic distributions are projected onto a 2D plane, and a Fourier transform is computed to generate the diffraction patterns. This dataset is primarily used for machine learning and deep learning applications in scattering, imaging and microscopy research, aiming to address the challenge of fast real-time inference.

搜集汇总

背景与挑战

背景概述

Molecular-MNIST数据集包含20000条数据，涵盖10种不同大小分子的结构变体，通过分子动力学轨迹的2D截面投影生成形状图像和衍射图案。该数据集专为机器学习和深度学习在散射、成像和显微镜研究中的应用设计，旨在支持快速实时推理问题的解决。