A Molecular-MNIST Dataset for Machine Learning Study on Diffraction Imaging and Microscopy

TL;DR

This paper introduces a Molecular-MNIST dataset comprising 10 molecule types with 2,000 variants each, designed to facilitate machine learning research in diffraction imaging and microscopy.

Contribution

The paper presents a novel, large-scale dataset for machine learning applications in molecular imaging, filling a gap in available benchmark data for scattering and microscopy studies.

Findings

Provides a new benchmark dataset for molecular imaging

Enables evaluation of ML algorithms on diffraction and microscopy data

Supports research in scattering and structural analysis

Abstract

An image dataset of 10 different size molecules, where each molecule has 2,000 structural variants, is generated from the 2D cross-sectional projection of Molecular Dynamics trajectories. The purpose of this dataset is to provide a benchmark dataset for the increasing need of machine learning, deep learning and image processing on the study of scattering, imaging and microscopy.