Image Dataset of Mango Leaf Disease for Deep Learning-Based Classification and Detection

The Original Mango Leaf Diseases Dataset is a carefully chosen collection of images that can be used to help researchers use computer vision and deep learning to automatically find and classify diseases on mango leaves. We took pictures of mango trees in the fields of Manikgonj and Savar, Bangladesh, using natural light and different backgrounds to make it look like real-world farming circumstances.

There are 2,364 original photos in the collection, which are split into six groups that show both healthy and sick mango leaves:
Anthracnose (326 pictures)
Bacterial Canker (524 pictures)
Gall Midge (207 pictures)
Healthy (316 pictures)
Powdery Mildew (414 pictures)
577 pictures of sooty mold

All of the pictures were taken in HEIC format and then changed to JPG (224×224 px) with the background removed (black background) so that they could all be processed the same way. To make the dataset more robust, data augmentation methods like rotation, flipping, brightness correction, and scaling were used. This made a balanced set of 6,000 images (1,000 each class).

The dataset is split into:
Set of 70% for training
20% of the validation set
10% of the test set

Use the dataset for:
1. Using CNNs to sort diseases into more than one class
2. Research on image segmentation and feature extraction
3. AI systems for farming and apps for finding diseases on the go

Format:
JPEG (RGB, 224×224 pixels) in folders with labels.