PubMed · 2026-06-29
Researchers built a large, freely available image dataset of shiitake mushrooms photographed in real growing conditions, covering three varieties and four categories including growth stages and deformed specimens. The dataset is designed to train AI systems that can automate mushroom grading, harvesting robots, and yield estimation.
The dataset includes 1,782 original high-resolution images expanded to 6,500 via augmentation, with 43,752 individually annotated mushroom instances.
Three commercially important varieties (9608, Chunsheng No. 1, Qihe No. 9) are represented across three growth stages plus a deformed-mushroom category.
Images capture realistic challenges including complex lighting, occlusion, dense clustering, and multiple viewing angles, making the dataset applicable to real-world AI deployment.
