Annotated Dataset for Potato Foliar Disease and Healthy Leaf Detection (Septoria, Early Blight, Late Blight, Healthy)

This dataset is strictly designed for scientific research in Artificial Intelligence, specifically for computer vision applications in phytopathology. It contains a total of 3,060 images with high-quality manual bounding box annotations for object detection tasks across four distinct classes: Septoria, Early Blight (Alternaria), Late Blight (Phytophthora infestans), and Healthy Leaves.

To create a robust and highly generalizable benchmark, this dataset is a strategic combination of two sources:
1. 908 novel field images collected in real-world conditions from potato crops in the Carchi province, Ecuador (comprising 548 exclusive images of Septoria, 231 Healthy, 72 Late Blight, and 57 Early Blight).
2. 2,152 laboratory images sourced from the public PlantVillage dataset (1,000 Early Blight, 1,000 Late Blight, and 152 Healthy) to establish a baseline.

Key scientific contribution: All 3,060 images, including the PlantVillage subset, were manually annotated with bounding boxes by the authors to enable Object Detection models, advancing beyond traditional image classification.

A precise preprocessing pipeline was applied: image resizing to 1024x1024 pixels (utilizing a "fit" strategy with white padding to prevent geometric distortion) and Contrast Limited Adaptive Histogram Equalization (CLAHE) for illumination standardization. No data augmentations (1x multiplier) have been applied. This repository serves as a reliable, transparent benchmark for training state-of-the-art models in precision agriculture.

Citation

If you use this dataset in your research or project, please cite the corresponding publication:

Pusdá-Chulde, M., Cuaycal-Tirira, D., Caranqui-Sánchez, V., García-Santillán, I. (2026). Detection of Leaf Diseases in Potato Crops using a lightweight Deep Learning model for Mobile Devices. Symmetry, 18(10), 1652. https://doi.org/10.3390/sym18101652