Rice Seed Detection and Counting Method

This dataset is developed for training rice seed recognition models, with the research hypothesis that color images of randomly distributed rice seeds captured by a standardized industrial camera can provide effective features for model training. The dataset contains color images of randomly distributed rice seeds, acquired using a HIKROBOT MV-CS060-10GC area scan industrial camera equipped with a Sony IMX178 CMOS sensor (1/1.8-inch sensor size) and featuring 6-megapixel resolution, rolling shutter technology, and a GigE data interface. Capable of color imaging, the camera achieves a maximum frame rate of 19.1 fps in Bayer RG8 format, with the lower surface of its lens positioned approximately 26 cm above the base plate, a focal length of 8 mm, a maximum aperture of F2.8, and a matching 1/1.8-inch imaging sensor size, along with an exposure time of 0.08 seconds. The data reflects clear imaging of randomly distributed rice seeds enabled by the camera’s calibrated parameters, ensuring the preservation of key visual features of the seeds. To interpret and use the dataset, users can directly apply it to model training for rice seed recognition after basic image preprocessing (e.g., format conversion), as the standardized data collection setup (fixed camera parameters and shooting distance) ensures consistency and usability, allowing others to replicate the imaging conditions and utilize the data for relevant model development.