Dataset of digital microscopy images of gelatin/siloxane 3D-printed lattice constructs for image-based quality assessment

This dataset contains 1000 JPEG digital microscopy images of gelatin/siloxane lattice constructs fabricated by extrusion-based three-dimensional printing. The images were acquired on a dark background to enhance contrast between the printed material and the surrounding field of view. The dataset was generated to support public reuse in image-based quality assessment, computer vision, and machine learning workflows applied to soft-material 3D printing. The dataset follows a non-randomized full-factorial experimental design composed of two nozzle diameters, five extrusion pressures, and five printing speeds. The nozzle diameters were 250 µm and 210 µm. The extrusion pressures were 160, 170, 180, 190, and 200 kPa. The printing speeds were 5, 10, 15, 20, and 25 mm/s. This 2 × 5 × 5 design resulted in 50 base printing conditions, with 20 repetitions per condition, producing a total of 1000 microscopy images. The dataset includes the raw image files, an image-level metadata table, a 50-condition experimental design table, a data dictionary, and documentation for reuse. The metadata file includes image identifiers, file names, image dimensions, file format, color mode, file size, SHA-256 checksum, descriptive intensity metrics, base run order, repetition number, nozzle diameter, extrusion pressure, and printing speed. These data can be reused for automated inspection of printed lattice constructs, image preprocessing, segmentation, filament continuity analysis, geometric fidelity assessment, similarity learning using Siamese neural networks, and convolutional neural network workflows such as ResNet50-based classification. The dataset is also intended to support related research articles focused on machine learning-assisted evaluation of gelatin/siloxane 3D-printed constructs.本数据集包含1000张通过挤出式三维打印制备的明胶/硅氧烷晶格结构体的JPEG格式数码显微图像。所有图像均以深色背景采集，以增强打印材料与周围视场之间的对比度。本数据集的构建旨在支持基于图像的质量评估、计算机视觉以及应用于软材料三维打印的机器学习流程等场景下的公开复用。 本数据集采用非随机化全因子实验设计，涵盖2种喷嘴直径、5种挤出压力与5种打印速度三类实验变量。其中喷嘴直径分别为250 µm与210 µm；挤出压力取值为160、170、180、190及200 kPa；打印速度取值为5、10、15、20及25 mm/s。该2×5×5的实验设计共生成50种基础打印工况，每种工况重复20次，最终总计得到1000张显微图像。 本数据集包含原始图像文件、图像级元数据表、50种工况的实验设计表、数据字典以及复用指南文档。元数据文件涵盖图像标识符、文件名、图像尺寸、文件格式、色彩模式、文件大小、SHA-256校验和、强度描述性指标、基础运行顺序、重复次数、喷嘴直径、挤出压力以及打印速度等信息。 本数据集可复用于打印晶格结构体的自动检测、图像预处理、图像分割、丝材连续性分析、几何保真度评估、基于孪生神经网络(Siamese neural networks)的相似度学习，以及基于ResNet50分类等卷积神经网络工作流。此外，本数据集还可支撑以明胶/硅氧烷三维打印结构体的机器学习辅助评估为主题的相关研究论文。