AI-Based Pest Detection Using the LeAF Dataset: A Vision for Supporting Smallholder Farmers
Students & Supervisors
Student Authors
Supervisors
Abstract
Pest infestation remains one of the most significant constraints to crop productivity for smallholder farmers, who often rely on manual scouting and visual inspection to identify early signs of damage. These traditional methods are time-consuming, inconsistent, and can delay timely decision-making. With advancements in precision agriculture, artificial intelligence (AI) offers a promising pathway for enhancing pest detection accuracy and accessibility. This study proposes the development of an AI-driven pest detection framework using the LeAF Pest Detection Dataset, a large-scale and publicly available dataset containing thousands of annotated pest images collected under realistic field conditions.The primary aim of this research is to explore how machine learning models trained on the LeAF dataset can support automated and efficient pest identification in smallholder farming environments. The study focuses on building and evaluating computer vision models capable of detecting multiple pest species across diverse leaf textures, lighting variations, and environmental contexts. Rather than presenting quantified results, this abstract outlines the intended goals: to assess the suitability of the dataset for real-world applications, examine model robustness across heterogeneous conditions, and evaluate how such systems could ultimately enhance early pest warning capabilities. In the full paper, we will investigate model performance, feature extraction methods, and deployment possibilities such as mobile applications for farmers. Expected outcomes include improved detection consistency, reduced dependency on expert intervention, and better-informed pest management strategies. By leveraging an open-source dataset, this study aims to contribute to accessible, scalable, and farmer-friendly AI solutions that support agricultural sustainability and resilience. The results, when completed, are intended to inform future development of AI-based tools that empower smallholder farmers with timely, accurate, and cost-effective pest monitoring technologies.
Keywords
Publication Details
- Type of Publication:
- Conference Name: International Conference on Regenerative Agriculture and Sustainable Food Security
- Date of Conference: 12/12/2025 - 12/12/2025
- Venue: Gazipur Agriculture University, Gazipur
- Organizer: Gazipur Agricultural University