Comprehensive Investigation of Machine Learning and Deep Learning Networks for Identifying Multispecies Tomato Insect...

Area of Science:

• Agricultural Science
• Computer Science
• Machine Learning

Background:

• Tomato crops (Solanum lycopersicum) are vulnerable to pests and drought, causing significant yield and financial losses.
• Accurate pest detection is vital for integrated pest management and sustainable agriculture.
• Current methods often lack the efficiency and accuracy needed for large-scale agricultural applications.

Purpose of the Study:

• To explore the effectiveness of Convolutional Neural Networks (CNNs) for automatic tomato pest image classification.
• To investigate and compare the performance of various optimizers (AdaDelta, AdaGrad, Adam, RMSprop, SGD, Nadam) on CNN-based pest classification.
• To evaluate the optimized CNN models against conventional machine learning algorithms and state-of-the-art deep learning models.

Main Methods:

• A customized CNN model was trained and evaluated on a dataset of 4263 tomato pest images.
• The performance of six different optimizers was compared based on classification accuracy, convergence speed, and robustness.
• Conventional machine learning models (logistic regression, random forest, naive Bayes, SVM, decision tree, KNN) were used as benchmarks.

Main Results:

• RMSprop achieved the highest validation accuracy (89.09%) among individual optimizers, with strong precision, recall, and F1 scores.
• Cross-validation revealed Nadam optimizer with CNN outperformed other approaches, achieving a mean accuracy of 79.12% and F1 score of 78.92%.
• Optimized CNN approaches demonstrated superior performance compared to conventional machine learning models and other deep learning architectures.

Conclusions:

• Specific optimizers significantly impact CNN performance for tomato pest classification.
• Nadam and RMSprop-optimized CNNs offer effective solutions for automated pest detection in tomato cultivation.
• This research provides valuable guidance for agricultural image analysis and enhances automated pest management strategies.