This paper presents a dataset with 5513 images of individual soybean seeds, which encompass five categories: (Ⅰ) Intact, (Ⅱ) Immature, (Ⅲ) Skin-damaged, (Ⅳ) Spotted, and (Ⅴ) Broken. Furthermore, there are over 1000 images of soybean seeds in each category. Those images of individual soybeans were classified into five categories based on the Standard of Soybean Classification (GB1352-2009) [1]. The soybean images with the seeds in physical touch were captured by an industrial camera. Subsequently, individual soybean images (227×227 pixels) were divided from the soybean images (3072×2048 pixels) using an image-processing algorithm with a segmentation accuracy of over 98%. The dataset can serve to study the classification or quality assessment of soybean seeds.
