WheatVision: A Dataset of Wheat Seed Quality Assessment

Description

This dataset presents a curated collection of high-resolution wheat seed images developed to advance automated quality assessment in agricultural grain inspection. Captured at the individual seed level, the dataset enables fine-grained classification between healthy seeds and multiple categories of physical defects, including insect infestation, mechanical damage such as cracks and dents, surface punctures, discoloration patterns like black point and staining, mold-affected seeds, and shriveled or underdeveloped grains. Designed with real-world post-harvest and storage inspection challenges in mind, this dataset supports the development of computer vision and deep learning models for tasks such as seed sorting automation, defect detection, and grain quality grading. Each image is precisely annotated to reflect the specific condition of the seed, offering a reliable foundation for both classification and segmentation-based research. By capturing the diversity of defects that commonly affect wheat quality during storage and handling, this dataset aims to bridge the gap between traditional manual inspection methods and scalable, AI-driven grain quality control systems — contributing toward more efficient, accurate, and non-destructive evaluation techniques in precision agriculture.

Steps to reproduce

Wheat seeds were procured from a local grain market and categorized into healthy and defective classes based on visual inspection. Images were captured using a smartphone camera against a plain white background under consistent natural lighting, followed by quality verification, annotation, and organization into class-specific folders. The dataset can be reproduced by following the same seed collection, image acquisition, and labeling protocol.

Institutions

• Vishwakarma Institute of Technology

  Maharashtra, Pune

Categories

Licence