ParasitoBank: Dataset of helminths and protozoa in coprology samples

This dataset contains 779 images of the visual field from fresh fecal samples analyzed under the microscope. These images were acquired using a Motorola G84 cellphone at resolutions of 3,072 x 3,072 pixels and 3,072 x 4,080 pixels. Parasite content was labeled by outlining regions, resulting in a total of 1,620 intestinal parasites, with a particular focus on intestinal protozoa. Through image processing, non-informative and noisy regions that did not provide relevant information within the visual field were removed, yielding images with an approximately 1:1 aspect ratio and a resolution of 2,100 x 2,100 pixels. The labeling information was organized in a JSON file following the “Common Objects in Context” (COCO) format, which was combined with the folder containing the images into a compressed file.

本数据集包含779张显微镜下新鲜粪便样本视野的显微图像。所有图像均通过摩托罗拉（Motorola）G84手机采集，原始分辨率分别为3072×3072像素与3072×4080像素。研究人员通过轮廓标注的方式对图像中的寄生虫区域进行标注，共计得到1620个肠道寄生虫标注实例，研究重点尤其聚焦于肠道原生动物。经图像处理流程，我们剔除了视野内无有效参考价值的冗余区域与噪声区块，最终生成宽高比约为1:1、分辨率为2100×2100像素的标准化图像。标注信息以JSON文件形式存储，并遵循通用目标检测数据集（Common Objects in Context, COCO）格式规范，最终将该标注文件与图像文件夹打包为压缩归档文件。