Manalagi Apple Disease Dataset

Description

Purpose Classification and identification of Manalagi apple fruit diseases Type of data Image files. Data format Joint Photographic Experts Group (JPG). Image resolution 1024 × 1024 pixels. Number of classes Four: - Healthy - Anthracnose - Black Pox - Powdery Mildew Dataset structure Raw_Images: 5,100 images Original_Images: 482 images Augmented_Images: 4,718 images Metadata The dataset contains labeled images categorized into four classes of Manalagi apple fruit conditions. The raw image folder contains the original images collected directly from orchards. The processed image folder includes a curated dataset divided into training, validation, and testing subsets, together with augmented training images. Images were captured under natural orchard conditions with variations in illumination, background, viewing angle, and disease severity to represent real-world agricultural environments. Data acquisition Images were collected directly from farmer-owned apple orchards using digital cameras and smartphone cameras under natural environmental conditions. The acquisition process was designed to capture real-world variability in fruit appearance and disease symptoms. Data source Agricultural source Orchard location: Poncokusumo, Malang Regency, East Java, Indonesia Source: Local farmer-owned apple orchards

Steps to reproduce

1. Data Collection Collect images of Manalagi apple fruits from farmer-owned orchards in Poncokusumo, Malang Regency, East Java, Indonesia. Capture both healthy and diseased fruits under natural orchard conditions. 2. Field Verification Consult local farmers and agricultural experts to verify disease symptoms and define the four disease categories. 3. Image Acquisition Capture images using smartphone and digital cameras with variations in viewing angle, distance, lighting conditions, and background to reflect real-world environments. 4. Data Validation and Labeling Review and label the collected images into four categories: Healthy Anthracnose Black Pox Powdery Mildew 5. Data Preprocessing Resize all selected images to a uniform resolution of 1024 × 1024 pixels, convert them to JPG format, and organize them into class-specific folders. Split the curated dataset into training, validation, and testing subsets. 6. Data Augmentation Apply augmentation techniques to the training images, including brightness adjustment, horizontal flipping, rotation, and scaling, to improve dataset diversity while preserving class labels. 7. Final Dataset Preparation Prepare the final dataset with the following structure: Raw Images: 5,100 images Processed Original Images: 482 images (Train, Validation, and Test) Processed Augmented Images: 4,718 images

Institutions

• Bina Nusantara University

  DKI Jakarta, West Jakarta

Categories

Licence