Detection of Potato Plant Leaf Diseases using Convolutional Neural Networks
A 1.7 MB CNN classifies potato leaf blight, reaching 98.66% on PlantVillage and 90.28% in Ethiopian fields.
Researchers trained a lightweight CNN to classify potato leaves as healthy, early blight, or late blight, targeting foliar disease that affects production in Ethiopia. Using 2,215 PlantVillage images, it reached 99.04% training, 98.21% validation, and 98.66% test accuracy, and 90.28% on 247 independent field images from Banja, Awi Zone. The model has 446,787 trainable parameters, is about 1.70 MB, and averaged 5.58 seconds per epoch versus 9.96 for VGG16 and 7.68 for MobileNetV2. The authors emphasize field evaluation under more variable imaging conditions.
- A compact CNN classifies healthy, early blight, and late blight potato leaves.
- PlantVillage test accuracy was 98.66%; an Ethiopian field set scored 90.28%.
- The model has 446,787 parameters and is about 1.70 MB.
- It trains faster per epoch than VGG16 and MobileNetV2 in the reported setup.
- Countries
- Ethiopia
- AI models
- MobileNetV2VGG16
From the original article
1 min read at arxiv.orgPotato is an important food-security crop in Ethiopia, but its production is affected by foliar diseases, particularly early blight and late blight. This study develops a lightweight convolutional neural network (CNN) to classify potato leaves into three categories: healthy, early blight, and late blight.
By Haimanot Bitew Alemayehu · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.