Annotated images of microscopy images from carbon fibre composites

Four data sets containing 500, 500, 1000 and 32 images of 256 x 256 pixels of parts of microscopy images of cross sections of carbon fibre composite laminates. The data sets contain corresponding annotated masks for every image. Each pixel is either annotated in the classes: matrix material, fibre or void and is respectively labelled with 0, 1 and 2. The masks are made by locating the centres of the fibres with a find local maxima method in ImageJ. These locations are then used to draw perfect circles with the average fibre radius around them to indicate the fibres. The voids are detected using a thresholding algorithm based on a minimum method. The data sets can be used to train, validate and test semantic segmentation algorithms.<br><br>

本数据集包含4组样本，分别包含500、500、1000与32张分辨率为256×256像素的碳纤维复合材料层合板横截面显微图像的局部区域图像。所有图像均配有对应的标注掩码（mask），每张图像的每个像素均被归入以下三类之一：基体材料、纤维或孔隙，分别对应标签0、1与2。<br><br>掩码的生成流程为：首先在ImageJ软件中通过局部极大值查找法定位纤维中心，随后以这些纤维中心为圆心、以平均纤维半径绘制正圆以标注纤维区域；孔隙区域则通过基于最小值法的阈值分割算法完成检测与标注。本数据集可用于训练、验证与测试语义分割算法。