Early detection of the Taro Leaf Blight disease in the West African sub-region using deep image classification models

Abstract

Taro, a vital crop in West Africa, is under attack by a disease called Taro Leaf Blight, which is bad news for the economy and farmer since it severely affects their income. Our study tackles the tough parts of spotting plant diseases, like the need for diverse datasets and better ways to analyze images. We're focusing on making a top-notch dataset just for West Africa and using some of the latest tech in deep learning—like VGG16, ResNet50, InceptionV3, MobileNetV2, DenseNet121, Xception, and the Vision Transformer—to spot the disease. From our research, it turns out, the Vision Transformer is the star player here, nailing a 74% success rate in picking out the disease in images, which is way better than older methods like VGG16 and ResNet50 that scored 56% and 36%. In addition, we're digging into how this can help manage diseases in West African farms, facing current problems head-on and suggesting new ways to make things better, not just for Taro but other crops too. The findings are a win-win for tech and farming, offering solid plans to fight back against this blight and keep crops healthy.

Keywords

• Taro Leaf Blight
• Deep Learning Models
• Image Classification Techniques
• Vision Transformer
• West African Agriculture
• Plant Disease Detection