PakRice_V1

This repository contains a structured rice-grain image dataset and associated research assets for computer-vision-based rice variety classification, grain segmentation, morphological feature analysis, and deployment. The archive is organized into three main sections: Data_Set, Cloud, and Diagrams, and contains 524 directories and 3,064 files in total. In addition to image data, the repository includes web-application code, serialized models, algorithm diagrams, and supporting deployment assets, making it suitable as an end-to-end resource for rice-image analysis and agricultural AI research. The Data_Set branch is divided into two major image collections: Scale_paper and Scanner Dataset. The Scale_paper collection represents scale-chart / paper-based image samples, where grains appear to be captured on a controlled background for measurement-oriented analysis. In the visible folder structure, this branch includes varieties such as C9 Sella, PK 386, Super 109, Super Brown, Super Silky, Super White, and Supri Sella, largely organized under non-connected rice subsets. This branch is therefore well suited for experiments involving grain measurement, visual feature extraction, and morphology-aware classification under controlled image-acquisition settings. The Scanner Dataset represents scanner-based rice images, captured in a more standardized flatbed scanning environment. This branch includes rice varieties such as Basmati2000, C9_Sella, Kainat1121, Kainat1121Sella, PK_386, Sella1509, and Super_109, with further subdivisions describing grain presentation conditions, including Connected_Rice, Non_connected_Rice, Overlaped, and Single. These subset labels indicate whether grains are touching, separated, overlapping, or individually isolated, which makes the dataset useful for classification, segmentation, instance separation, and robustness testing across different sample arrangements.