Image Dataset of Local Cocoa, Maize and Soybean Seed Varieties from Ghana

This dataset comprises 31,069 confirmed, expert-verified single-seed images spanning 12 locally recognised varieties across three economically important Ghanaian crops: cocoa (4 varieties—Agric, Gardener, Hybrid, and Tetteh Quarshie; 3,005 images), maize (4 varieties—Bhihilifa, Obatanpa, Sanzal Sima, Wang Basik; 13,438 images), and soyabean (4 varieties—Afayak, FarmerSaved, Favor, and Jenguma; 14,626 images). Maize and soyabean seeds were pre-sorted by variety at Heritage Seeds Ghana; cocoa seeds were extracted from pods pre-sorted by variety by farm experts in Berekum, Ghana, and sun-dried for seven days before imaging. All seeds were photographed in 352 groups against a plain blue background with a Google Pixel 9a smartphone under natural daylight, then automatically cropped into individual seed images (native 640 × 640 px) using a colour-based segmentation pipeline, with variety labels independently re-confirmed by the respective experts — Heritage Seeds Ghana for maize and soyabean, and the Quality Control unit of the Ghana Cocoa Board (COCOBOD), Sunyani, alongside Berekum farm experts, for cocoa—before every cropped image underwent a further manual visual audit. Images are released at their native 640 × 640 px resolution and organized into folders by crop and variety. The dataset supports the development of low-cost, mobile deep-learning models for seed variety identification and certification, addressing the current lack of openly available, cultivar-labeled image data for African seed crops. Class distributions are close to balanced, though a few varieties—Agric cocoa (17.7%) and Bhihilifa maize (20.2%) among them—are comparatively under-represented, while Wang Basik maize (34.3%) is comparatively over-represented.