Research Area: Machine Learning
Plant diseases are a vital risk to crop yield and early detection of plant diseases remains a complex task for the farmers due to the similar appearance in color, shape, and texture. In this work, authors have proposed an automatic plant disease detection technique using deep ensemble neural networks (DENN). Transfer learning is employed to fine-tune the pre-trained models. Data augmentation techniques include image enhancement, rotation, scaling, and translation are applied to overcome overfitting. This paper presents a detailed taxonomy on the performance of different pre-trained neural networks and presents the performance of a weighted ensemble of those models relevant to plant leaf disease detection. Further, the performance of the proposed work is evaluated on publicly available plant village dataset, which comprises of 38 classes collected from 14 crops. The performance of DENN outperform state-of-the-art pre-trained models such as ResNet 50 & 101, InceptionV3, DenseNet 121 & 201, MobileNetV3, and NasNet. Performance evaluation of the proposed model demonstrates that effective in categorizing various types of plant diseases that comparatively outperform pre-trained models.
Keywords:
Transfer Learning
Deep Ensemble Neural Network
Plant Leaf Disease Detection
Deep Learning
Machine Learning
Author(s) Name: Sasikala Vallabhajosyula, Venkatramaphanikumar Sistla & Venkata Krishna Kishore Kolli
Journal name: Journal of Plant Diseases and Protection
Conferrence name:
Publisher name: Springer
DOI: 10.1007/s41348-021-00465-8
Volume Information: volume 129, pages545–558 (2022)
