A Deep Learning Approach to Efficient Corn Pest Detection for Sustainable Agriculture

Overview

Corn pests cause hundreds of millions of dollars in annual crop losses, yet current pest management relies on broad-spectrum chemical pesticides that harm ecosystems, beneficial insects, and human health. This project developed an AI-powered computer vision system to enable targeted, sustainable pest control through accurate real-time species identification. Using the IP102 benchmark dataset, I trained YOLOv8 and YOLOv9 object detection models on 18,976 images across 11 corn pest species. Through 50 systematic experiments with varying learning rates (0.0001-0.01) and training epochs (10-50), I optimized model performance using mean Average Precision at 50% Intersection over Union (mAP@50) as the primary evaluation metric. YOLOv9 achieved superior performance with 75.2% validation mAP@50 and 69.3% test mAP@50, compared to YOLOv8's 73.4% validation and 68.9% test accuracy. Per-class analysis revealed a critical challenge: three species, specifically white margined moth, large cutworm, and yellow cutworm, achieved only 14-33% accuracy due to severe class imbalance and visual similarity. I addressed this through targeted data augmentation specifically for underperforming classes. Yellow cutworm's accuracy improved the most from 33.5% to 76.3%. An ensemble model combining the base YOLOv9 with the augmentation-focused model achieved 75.7% test mAP@50, the highest result in the study. The system was deployed as a web application enabling real-time pest detection from uploaded images. This work demonstrates that deep learning can provide accurate, scalable solutions for agricultural pest monitoring, offering a practical pathway toward reducing chemical pesticide dependency and advancing precision agriculture.

Competition history

• CSEF 2026 Computational Science (Senior Division) · Entry S-07-23

Related projects

Closest projects by meaning, across every fair and year in the corpus.

Browse more like this

• All Computational Sciences projects
• CSEF Computational Sciences projects
• Every CSEF 2026 project