BDPapayaDisease: A Multi-Class Papaya Leaf Disease and Pest Image Dataset with Raw, Resized, and Color Metadata

BDPapayaDisease is a field-acquired smartphone image dataset developed for automated recognition of papaya (Carica papaya L.) leaf diseases and pest infestation under real-world agricultural conditions in Bangladesh. The dataset contains 4,506 JPEG images organized into six classes: Anthracnose (611 images), Black Spot (565), Leaf Curl Disease (740), Mealybug Infestation (871), Mosaic Virus (657), and Healthy Leaf (1,062). The images were collected between 15 June and 30 July 2026 from papaya-growing areas in five districts of Bangladesh: Dhaka, Chattogram, Sirajganj, Jamalpur, and Habiganj. Images were captured under natural daylight using six consumer-grade smartphones and under diverse field conditions, including different illumination, viewing angles, leaf orientations, backgrounds, and weather conditions. Both attached and detached leaves were photographed. Detached leaves were placed on backgrounds such as white A4 paper, black sheets, natural soil, and paved road surfaces to provide both relatively controlled and unconstrained imaging conditions. This Zenodo release contains both the high-resolution source image dataset (containing 6 individual files seperately) and a standardized 800 × 600 pixel version prepared for machine-learning and deep-learning applications. The images are organized into class-specific folders. No synthetic data augmentation was applied; however, a limited number of Mosaic Virus samples were generated by cropping high-resolution source photographs to represent spatially distinct symptomatic regions, followed by manual visual inspection. The repository also includes four supporting CSV files: dataset_specification.csv, containing image-level characteristics and quality information; comprehensive_statistics.csv, containing class-wise descriptive statistics; mean_rgb_values.csv, containing class-level mean RGB color values; and color_similarity_matrix.csv, describing inter-class color similarity based on normalized Euclidean-distance analysis. The image-quality assessment includes information related to resolution, file size, megapixels, aspect ratio, sharpness, brightness, contrast, and estimated noise level. The dataset is intended to support research in papaya leaf disease classification, pest-infestation recognition, computer vision, machine learning, deep learning, precision agriculture, image-quality analysis, and mobile-based plant disease diagnosis. It provides realistic visual variability that can be useful for developing, training, validating, and comparatively evaluating image-based disease-recognition models.