[Paper Review] Unleashing the Power of Transfer Learning Model for Sophisticated Insect Detection: Revolutionizing Insect Classification

Md. Mahmudul Hasan, SM Shaqib+4 more|arXiv (Cornell University)|Jun 11, 2024

Smart Agriculture and AIAgricultural and Biological Sciences3 citations

TL;DR

This study proposes a transfer learning-based insect detection system using deep CNNs, with ResNet152V2 achieving 97% test accuracy on insect classification. By leveraging pre-trained models on image datasets and fine-tuning on a custom insect dataset, the system enables high-accuracy, real-time detection for crop and plant health monitoring, significantly improving early pest identification in agriculture.

ABSTRACT

The purpose of the Insect Detection System for Crop and Plant Health is to keep an eye out for and identify insect infestations in farming areas. By utilizing cutting-edge technology like computer vision and machine learning, the system seeks to identify hazardous insects early and accurately. This would enable prompt response to save crops and maintain optimal plant health. The Method of this study includes Data Acquisition, Preprocessing, Data splitting, Model Implementation and Model evaluation. Different models like MobileNetV2, ResNet152V2, Xecption, Custom CNN was used in this study. In order to categorize insect photos, a Convolutional Neural Network (CNN) based on the ResNet152V2 architecture is constructed and evaluated in this work. Achieving 99% training accuracy and 97% testing accuracy, ResNet152V2 demonstrates superior performance among four implemented models. The results highlight its potential for real-world applications in insect classification and entomology studies, emphasizing efficiency and accuracy. To ensure food security and sustain agricultural output globally, finding insects is crucial. Cutting-edge technology, such as ResNet152V2 models, greatly influence automating and improving the accuracy of insect identification. Efficient insect detection not only minimizes crop losses but also enhances agricultural productivity, contributing to sustainable food production. This underscores the pivotal role of technology in addressing challenges related to global food security.

Motivation & Objective

• To develop an accurate, automated insect detection system for early identification of pest infestations in agricultural settings.
• To evaluate the performance of multiple transfer learning models, including MobileNetV2, ResNet152V2, Xception, and a custom CNN, on insect image classification.
• To demonstrate the feasibility of using deep learning with transfer learning for real-world agricultural applications in entomology and crop health monitoring.
• To contribute to global food security by minimizing crop losses through early and precise insect detection.

Proposed method

• Data acquisition from diverse agricultural sources to build a comprehensive insect image dataset.
• Image preprocessing steps including resizing, normalization, and data augmentation to improve model generalization.
• Splitting the dataset into training (80%) and testing (20%) sets to ensure unbiased evaluation.
• Implementation of transfer learning using pre-trained models—ResNet152V2, MobileNetV2, Xception, and a custom CNN—fine-tuned on the insect dataset.
• Training the models using stochastic gradient descent with categorical cross-entropy loss and early stopping to prevent overfitting.
• Evaluation of model performance using accuracy, precision, recall, and F1-score on the test set.

Experimental results

Research questions

• RQ1 Which transfer learning model achieves the highest accuracy in classifying insect species from agricultural images?
• RQ2 How does transfer learning improve performance on small or specialized insect datasets compared to training from scratch?
• RQ3 Can ResNet152V2 outperform other state-of-the-art architectures in insect detection tasks?
• RQ4 To what extent does data augmentation and preprocessing enhance model generalization in insect classification?

Key findings

• ResNet152V2 achieved 99% training accuracy and 97% testing accuracy, outperforming MobileNetV2, Xception, and the custom CNN.
• The high test accuracy of 97% indicates strong generalization and robustness to unseen insect images in real-world settings.
• Transfer learning significantly improved model performance on limited insect datasets, reducing training time and data requirements.
• The results confirm that pre-trained models like ResNet152V2 are highly effective for fine-grained insect classification tasks.
• The system demonstrates strong potential for deployment in precision agriculture for early pest detection and crop protection.