Dataset of digital microscopy images of gelatin/siloxane 3D-printed lattice constructs for image-based quality assessment
Description
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.
Files
Steps to reproduce
Gelatin/siloxane lattice constructs were fabricated using extrusion-based three-dimensional printing. The dataset was generated using 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. The 2 × 5 × 5 factorial design generated 50 base printing conditions. Each condition was repeated 20 times, resulting in 1000 printed constructs and 1000 microscopy images. After fabrication, each construct was imaged using digital microscopy on a dark background to increase contrast between the gelatin/siloxane material and the surrounding field of view. The images were saved in JPEG format and named sequentially from 1.jpg to 1000.jpg. The accompanying metadata file was generated by extracting file-level information from each image, including image identifier, file name, image dimensions, color mode, file size, SHA-256 checksum, descriptive image-intensity metrics, base run order, repetition number, nozzle diameter, extrusion pressure, and printing speed.
Institutions
- Universidad Autónoma de Ciudad JuárezChihuahua, Ciudad Juárez