2026-10-06T22:05:53Z https://api.figshare.com/v2/oai

oai:figshare.com:article/34005402 2026-09-27T19:37:42Z category_24064 category_28885 item_type_3 month_year_09_2026

<b>AgroVisionBD:</b> A Primary Image Dataset of Healthy and Diseased Banana, Mango, and Papaya Fruits and Leaves from Bangladesh Rafid Haque (25116276) Horticultural crop protection (incl. pests, diseases and weeds) Image processing AgroVisionBD Plant disease Fruit disease Leaf disease Plant pathology agricultural computer vision <p dir="ltr"><b>AgroVisionBD</b> is a primary image collection of healthy and diseased banana, mango, and papaya fruits and leaves acquired in Bangladesh. The repository contains 5,280 RGB JPEG images organized into 18 classes, comprising 12 disease-related and six healthy classes. The collection includes 2,261 fruit images and 3,019 leaf images.</p><p dir="ltr">Images were collected between May and September 2025 from Mirpur, Dhaka, and Patuakhali, Barishal Division, Bangladesh. Image acquisition was performed using Apple iPhone 16 Pro Max and Xiaomi Redmi Note 15 Pro smartphones under practical acquisition conditions. Variations in background, illumination, viewing angle, orientation, acquisition distance, and visible disease severity were retained. All images included in this release were collected directly by the research team; no images obtained from websites or previously published public datasets are included.</p><p dir="ltr">Images were manually curated for visual quality and class consistency, while perceptual hashing (pHash) was used to support duplicate and near-duplicate screening. An external academic expert in agronomy reviewed all 18 class definitions together with representative images. Sixteen classes were recorded as Accepted, while <code>mango_anthracnose_fruit</code> and <code>mango_black_spot_fruit</code> were recorded as Moderate because of overlapping visual characteristics. The validation represents class-level visual assessment and does not constitute laboratory or pathogen-confirmed diagnosis.</p><p dir="ltr">The repository includes the original-resolution image collection, image-level <code>metadata.csv</code>, class-level documentation, expert-validation information, and a README describing the dataset structure and reuse conditions. Image-level metadata include class information, collection location, capture device, image dimensions, image format, color mode, file size, and relative repository path.</p><p dir="ltr">AgroVisionBD is intended for reuse in plant disease classification, transfer learning, agricultural computer vision, machine learning, feature extraction, explainable AI, class-imbalance analysis, and related image-based agricultural research.</p> 2026-09-27T19:37:42Z Dataset Dataset 10.6084/m9.figshare.34005402.v1 https://figshare.com/articles/dataset/_b_AgroVisionBD_b_A_Primary_Image_Dataset_of_Healthy_and_Diseased_Banana_Mango_and_Papaya_Fruits_and_Leaves_from_Bangladesh/34005402 CC BY 4.0