Pertanian & Lingkungan editorial
Open AccessOA2023
EffiMob-Net: A Deep Learning-Based Hybrid Model for Detection and Identification of Tomato Diseases Using Leaf Images
EffiMob-Net, a hybrid deep learning model combining EfficientNetB3 and MobileNet, detects tomato leaf diseases with 99.92% accuracy on the PlantVillage dataset, offering a reliable automatic detector for agricultural use.
Zahid Ullah; Najah M. Alsubaie; M. Jamjoom; S. Alajmani; F. Saleem· Agriculture· 2023· DOI 10.3390/agriculture13030737
The core problem
Tomatoes are the most consumed vegetable globally, and their production is threatened by pests and diseases exacerbated by global warming and climate change. Detecting and preventing such diseases is crucial for ensuring tomato plant survival and yield. Artificial intelligence (AI), particularly deep learning (DL), has recently contributed significantly to agronomy by enabling disease detection through leaf images. This paper proposes a hybrid DL-based approach, EffiMob-Net, which fuses two pretrained models—EfficientNetB3 and MobileNet—to accurately detect tomato leaf diseases. The study addresses the need for reliable automatic detectors to support agricultural productivity.
Innovation
The proposed EffiMob-Net model is a fusion of EfficientNetB3 and MobileNet architectures. To mitigate overfitting, the authors employed regularization, dropout, and batch normalization (BN). Hyperparameter tuning was conducted to select optimal parameters for building the best-fitting model. The model was trained and tested on the PlantVillage dataset, which contains images of tomato leaves both diseased and healthy. The hybrid architecture leverages the strengths of both base models: EfficientNetB3 for efficient feature extraction and MobileNet for lightweight computation. The fusion is designed to capture complementary features, enhancing detection accuracy. The model's performance was evaluated using accuracy metrics, with the best classifier selected for disease detection.
Introduction
Tomatoes are the most consumed vegetable globally, and their production is threatened by pests and diseases exacerbated by global warming and climate change. Detecting and preventing such diseases is crucial for ensuring tomato plant survival and yield. Artificial intelligence (AI), particularly deep learning (DL), has recently contributed significantly to agronomy by enabling disease detection through leaf images. This paper proposes a hybrid DL-based approach, EffiMob-Net, which fuses two pretrained models—EfficientNetB3 and MobileNet—to accurately detect tomato leaf diseases. The study addresses the need for reliable automatic detectors to support agricultural productivity.
Why it matters
The findings show that EffiMob-Net is a reliable automatic detector for tomato plant diseases, with significant potential to contribute to better solutions for detecting other crop diseases in agriculture. The high accuracy (99.92%) underscores the effectiveness of hybrid deep learning models in precision agriculture. The use of regularization, dropout, and batch normalization was crucial in preventing overfitting, ensuring the model generalizes well to unseen data. Hyperparameter tuning further optimized performance. The study's limitation includes evaluation on a single dataset (PlantVillage), which may not fully represent field conditions. Future work could extend the model to other crops and real-world scenarios. Overall, EffiMob-Net represents a promising advancement in AI-driven plant disease detection, offering a scalable and accurate tool for farmers and agronomists.
Who should read this
Pembaca Research Digest dan praktisi rumpun ini.
Opening member content…