Soybean Cultivar Classification Dataset

Published: 9 May 2026| Version 1 | DOI: 10.17632/snc7gjw632.1
Contributor:
jinchao yuan

Description

This dataset contains 11 types of soybean images from the rural area of Jilin Province's Jiuchu County, totaling 35,039 images: Beidou 47, Heihe 53, Heike 59, Heike 60, Heike 68, Heike 71, Heike 92, Huajiang Huinong 417, Jiadou 55, and Longken 3002.

Files

Steps to reproduce

We use a platform for acquiring soybean images. It includes a camera, an aluminium frame, black shading cloth, and a lighting system. The camera is mounted on the top of the smartphone's aluminium frame (VIVO V1986A) to maintain a consistent shooting angle. Two 50 W halogen lights form the illumination system. The process occurs in a shaded environment to reduce interference from ambient light. During image acquisition, seeds of the same variety were evenly placed and fixed on a counting plate, with approximately 100 seeds per image. After image capture, image segmentation is carried out by combining the image segmentation algorithm with manual assistance, which was segmented to extract individual seeds, yielding 35,039 samples.

Institutions

Categories

Computer Vision, Image Classification, Soybean

Licence