Ginseng Dataset For re-ID
Description
This dataset is designed for instance-level ginseng image retrieval and re-identification. Images are organized into one folder per physical ginseng instance. Images within the same folder depict the same instance under different placements and orientations, whereas images in different folders depict different instances. The dataset contains 1,075 high-resolution JPG images across 271 ginseng instances. Each instance is represented by two to four images, with most instances having four images. Common image resolutions include 4898 × 3288 and 4901 × 3288 pixels. These images preserve main-root contours, branch structures, and fine rootlets, supporting the evaluation of retrieval methods under substantial intra-instance variation. These images constitute the instance-labeled multi-view evaluation set in the associated ginseng re-identification study. They should be distinguished from the separate unlabeled image pool used for representation learning.