Synthetic datasets for colloidal particle assemblies
收藏资源简介:
该数据集由阿斯特拉罕塔季谢夫国立大学等机构创建,旨在支持胶体粒子二维组装的识别研究。数据集包含四种不同形状粒子(球形、椭球形、立方体和棒状)的合成图像,每个子集包含135至155张原始图像,通过旋转和添加噪声进行数据增强,并统一缩放至640x640像素,采用多边形标注进行实例分割。数据集通过人工生成,以解决实验显微图像稀缺的问题,专门用于训练YOLOv8模型,以自动化识别胶体组装中的孤立粒子、二聚体、链状、簇状和环状结构,应用于纳米技术、光子晶体和材料科学领域,旨在提升胶体结构形态分析的准确性和效率。
This dataset was created by Astrakhan Tatishchev State University and other institutions, aiming to support research on the recognition of two-dimensional assemblies of colloidal particles. The dataset includes synthetic images of particles with four distinct shapes: spherical, ellipsoidal, cubic, and rod-like. Each subset contains 135 to 155 raw images, which undergo data augmentation via rotation and noise addition, are uniformly resized to 640×640 pixels, and are annotated with polygons for instance segmentation. The dataset was artificially generated to address the scarcity of experimental microscopic images. It is specifically designed for training YOLOv8 models to automatically recognize isolated particles, dimers, chain-like, cluster-like, and ring-like structures in colloidal assemblies. Targeting applications in nanotechnology, photonic crystals, and materials science, this dataset aims to improve the accuracy and efficiency of morphological analysis for colloidal structures.




