A Comprehensive Review of Deep Learning Techniques for Crop Pest Detection

computer science and engineering

Abstract—Crop pest infestation remains one of the most significant threats to global food security, causing substantial eco-nomic losses and reducing agricultural productivity worldwide. Traditional manual inspection methods are labor-intensive, time-consuming, and often ineffective for early detection, particularly in large-scale farming operations. In recent years, deep learning has emerged as a transformative technology for automated pest detection, offering high accuracy, real-time processing, and scalability. This paper presents a comprehensive review of deep learning approaches for agricultural pest detection, covering object detection architectures (YOLO, Faster R-CNN, SSD), classification models (ResNet, EfficientNet, Vision Trans-formers), multimodal frameworks combining visual and textual data, and segmentation-based methods. We analyze state-of-the-art contributions including YOLOv8-based lightweight models, CNN-Transformer hybrids, IoT-integrated detection systems, and explainable AI techniques for pest recognition. Key datasets including Pest24, IP102, and Agricultural Pests Dataset are examined. We identify critical challenges including small object detection, class imbalance, similar species discrimination, and real-world deployment constraints. Finally, we present future research directions including few-shot learning, edge deployment, multimodal fusion, and integrated pest management systems with pesticide recommendation capabilities. This review serves as a comprehensive reference for researchers and practitioners developing intelligent pest monitoring solutions for precision agriculture.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.