Mango Fruit Image Classification - Uganda
Description
This dataset contains 37,957 high-resolution images of mango fruits collected in Eastern Uganda (Soroti District: Aloet and Madera areas) using smartphone cameras under natural daylight conditions. It includes both healthy and defective mangoes, representing a wide range of post-harvest conditions encountered during harvesting, handling, and marketing. Dataset Organization: Original: 4,699 raw images captured in the field. Preprocessed: 5,064 images resized and center-cropped for consistent framing. Augmented: 28,194 images generated via flipping, brightness adjustment, cropping, and controlled 90°/270° rotations to simulate natural orientation changes. File Format & Access: Images are stored in JPEG format. The dataset is provided in .zpaq format for maximum compression and can be extracted using PeaZip (Windows). A .zip version of the dataset is also available on Kaggle: https://www.kaggle.com/datasets/joanitanamuyiga/mango-fruit-image-classification-uganda Potential Applications: Computer vision and image processing research Fruit quality assessment and post-harvest defect studies Machine learning applications for classification, defect detection, and infection segmentation
Files
Steps to reproduce
1. Data Collection - Location: Aloet and Madera areas, Soroti District, Eastern Uganda. - Device: Smartphone cameras (various models). - Environment: Natural field and market conditions, daylight illumination. - Classes: Healthy (marketable) and defective (damaged/diseased) mango fruits. 2. Preprocessing - Images resized and center-cropped for uniform framing. - Maintained JPEG format to balance quality and storage. 3. Augmentation - Applied image transformations to expand dataset diversity: - Horizontal flipping. - Brightness adjustment. - Center Cropping - Controlled rotations at 90° and 270°. 4. Final Dataset - Total images: 37,957 (Original + Preprocessed + Augmented). - Structured into subfolders by class and stage (Healthy / Defective).