Plant leaves image segmentation dataset

Overview:

This repository contains high-resolution RGB images captured by field advisors and selected by experts for the dataset specifically designed for training two models to recognize Cercospora Leaf Spot Disease in beet and Brown Rust of Rye in rye. The dataset comprises 397 images and includes original photos along with binary masks for two categories: leaf segmentation and disease spot segmentation.

Images were labeled using the Label Studio tool with expertise from the Institute of Plant Protection in Poznań, Poland. The dataset has been utilized to train CNN models for product avaiable at AI4EOSC marketplace that are integrated into the mobile application of eDWIN advisory platform.

Dataset Composition:

Each image comes with corresponding masks that highlight the leaves and specific disease spots. The dataset is organized into separate directories for each plant and type of segmentation:

├── beet/│   ├── disease_segmentation/│   │   ├── images/│   │   │   │   ├── sick_12.jpg│   │   │   │   ├── sick_13.jpg│   │   │   │   ├── sick_15.jpg│   │   │   │   ├── sick_17.jpg│   │   │   │   ├── sick_18.jpg│   │   │   │   └── ...│   │   └── masks/│   │       │   ├── sick_12.png│   │       │   ├── sick_13.png│   │       │   ├── sick_15.png│   │       │   ├── sick_17.png│   │       │   ├── sick_18.png│   │       │   └── ...│   └── leaves_segmentation/│       ├── images/│       │   │   ├── 1.jpg│       │   │   ├── 10.JPG│       │   │   ├── 11.JPG│       │   │   ├── 12.JPG│       │   │   ├── 13.JPG│       │   │   └── ...│       └── masks/│           │   ├── 1.png│           │   ├── 10.png│           │   ├── 11.png│           │   ├── 12.png│           │   ├── 13.png│           │   └── ...└── rye/    ├── disease_segmentation/    │   ├── images/    │   │   │   ├── sick_0.jpg    │   │   │   ├── sick_102.jpg    │   │   │   ├── sick_103.jpg    │   │   │   ├── sick_104.jpg    │   │   │   ├── sick_105.jpg    │   │   │   └── ...    │   └── masks/    │       │   ├── sick_0.png    │       │   ├── sick_102.png    │       │   ├── sick_103.png    │       │   ├── sick_104.png    │       │   ├── sick_105.png    │       │   └── ...    └── leaves_segmentation/        ├── images/        │   │   ├── IMG_0809.JPG        │   │   ├── IMG_0810.JPG        │   │   ├── IMG_0811.JPG        │   │   ├── IMG_0812.JPG        │   │   ├── IMG_0813.JPG        │   │   └── ...        └── masks/            │   ├── IMG_0809.png            │   ├── IMG_0810.png            │   ├── IMG_0811.png            │   ├── IMG_0812.png            │   ├── IMG_0813.png            │   └── ...

Usage:

The dataset is structured to aid in the development and testing of models for agricultural disease detection. The following table represents structure of dataset_overview.csv:

Plant
Segmentation
Category
Filename

beet
disease_segmentation
images
sick_12.jpg

beet
disease_segmentation
masks
sick_12.jpg

rye
leaves_segmentation
images
sick_15.jpg

rye
leaves_segmentation
masks
sick_15.jpg

File Numbering/Naming Scheme:

The filenames for the images and masks correspond to the disease and leaf category they represent, with the RGB image names matching their respective binary masks. The naming convention is straightforward, reflecting the segmentation type and sequence number.

Note:

All data were recorded under various field conditions, and experts verified annotations to ensure the accuracy of labels. The dataset is continually updated with new data and annotations to enhance model robustness and accuracy.