VegQual : A Multiclass Dataset for Real-Time Fresh and Defective Vegetables

The dataset contains <b>2,032 raw images of seven commonly consumed vegetable species</b>, each categorized into <b>fresh and defective quality classes</b>, resulting in a total of <b>14 object categories</b>. The images were captured under real-world conditions and include variations in viewing angles, backgrounds, object distances, and lighting environments, providing a diverse and challenging resource for training and evaluating deep learning–based object detection models. Each image has been carefully annotated using bounding boxes in <b>TXT (YOLO) format</b>, which include class identifiers and normalized bounding box coordinates. These annotations enable precise object localization and are fully compatible with widely used deep learning frameworks. All annotations were created using the <b>Roboflow platform</b>, ensuring consistency, accuracy, and high-quality labeling standards.After annotation and preprocessing, the dataset contains <b>4,736 annotated instances across the 14 categories</b>. To support effective model training and evaluation, the dataset was divided into three subsets: <b>training, validation, and testing</b>. Specifically, <b>70% of the images (1,422 images)</b> were allocated for training, <b>20% (406 images)</b> for validation, and <b>10% (204 images)</b> for testing. Prior to training, all images were resized to a standard resolution of <b>640 × 640 pixels</b>, and automatic orientation correction was applied to address rotation inconsistencies caused by EXIF metadata.The dataset is organized into three primary directories <b>train</b>, <b>valid</b>, and <b>test </b>each containing <b>images/</b> and <b>labels/</b> subdirectories. Every image is associated with a corresponding label file that stores the class identifier and normalized bounding box coordinates for each detected object.Overall, the VegQual dataset provides a valuable benchmark for research in <b>computer vision, deep learning, agricultural automation, and food quality assessment</b>. It supports the development of robust models for <b>real-time vegetable classification and defect detection</b>, contributing to advancements in intelligent agricultural systems and sustainable food production.