Good and Bad Classification of Coriandrum sativum

Data Description: Good and Bad Classification of Coriandrum sativum

This dataset has been developed to support a classification project aiming to distinguish between "Good" and "Bad" quality samples of Coriandrum sativum (commonly known as coriander or cilantro). It contains a total of 2000 labeled images, evenly distributed between two categories:

1000 Good Samples: These images represent healthy, fresh coriander leaves. Characteristics typically include vibrant green coloration, full and intact leaf structures, absence of spots or wilting, and natural moisture. These samples meet visual standards for marketability and consumer acceptance.

1000 Bad Samples: These images depict coriander leaves that are not suitable for use or sale. They may show signs of spoilage such as yellowing, browning, black spots, mold presence, drying, shriveling, insect damage, or mechanical bruising. These defects affect both visual quality and shelf life.

Image Details:

Format: JPEG/PNG

Resolution: Varied (standardized during preprocessing)

Background: Mostly neutral or simple backgrounds to focus on leaf features

Capture Conditions: Mixed lighting and orientations to ensure dataset variability and robustness in model training

Purpose and Use: The dataset is intended for use in training and evaluating machine learning models, particularly in computer vision applications involving classification and quality assessment of agricultural produce. Potential applications include:

Automated sorting systems for coriander in supply chains

Mobile applications for farmers or vendors to assess produce quality

Research in plant disease and spoilage detection using AI

Labeling and Annotation: Each image is labeled as either "Good" or "Bad" based on visual inspection. Labeling was conducted by individuals with basic experience in agricultural product handling, ensuring practical relevance. No bounding boxes or segmentations are included; the task is a simple binary image classification.

Preprocessing Recommendations: To prepare the dataset for machine learning tasks, it is recommended to:

Resize images to a consistent resolution (e.g., 224x224 pixels for CNNs)

Normalize pixel values

Apply data augmentation (rotation, flipping, brightness variation) to improve model generalization

Ethical Considerations: This dataset is created solely for academic and research purposes. It is important to note that quality classification based on visual features may not always align perfectly with internal or chemical quality, and should be used as a complementary tool, not a sole decision-making source.