American
International
University-
Bangladesh

Students & Supervisors

Abstract

Agriculture is necessary for human survival, economic growth, and global food security. In this industry, outbreaks of pests and illnesses are still a big threat. They cause big yield losses and put the world's food supply at risk. Traditional identification methods are hard work, take a long time, and are likely to make mistakes because they often require personal inspection. Artificial intelligence (AI) is a branch of computer science that enables machines to perform tasks such as learning, reasoning, and problem-solving that often require human intelligence. AI has shown enormous potential in addressing these challenges. AI in farming can look at huge amounts of data from satellites, drones, and mobile imaging devices to make accurate predictions about possible outbreaks, find pest infestations, and identify sick crops. AI and deep learning technologies have become promising tools for accurate and speedy monitoring in recent years. They give farmers vital information that can help them manage their crops better, reduce losses, and improve food security overall. The main purpose of this study is to create and test an AI-based model that can detect important agricultural diseases and pests early. The study is focused on two particular studies: (a) How well are AI models like YOLOv8 at using drone and field- level pictures to discover diseases and damage done by pests? (b) How beneficial are these models in real life for farmers who need to make choices? The study uses a mixture of publicly accessible datasets (e.g., PlantVillage) and locally gathered field images of specific crops. We trained and refined deep learning models (YOLOv8 and transfer learning using CNNs like ResNet) on images that had been identified to show areas that were distressed or had pests. We used accuracy, precision, recall, F1-score, and mean average precision (mAP) to judge how well the model worked. In addition, field validation was conducted jointly with agricultural specialists to combine model predictions with scientific evaluations. The initial findings show that models based on YOLOv8 can find agricultural sickness and pest infestations with more than 85% accuracy. Also, lightweight versions of the model that were developed for edge devices (such as mobile phones and Jetson Nano) showed that it could be deployed in real time and at a minimal cost. This finding has important implications: prompt detection can cut down on the usage of pesticides that aren't needed, boost crop yields, and make farming more sustainable. AI-driven detection systems could give farmers more control, especially in areas with few resources, by giving them easy-to-use, dependable, and scalable tools for managing and protecting their crops.

Keywords

Publication Details

• Type of Publication:
• Conference Name: 11th INTERNATIONAL INSEARCH CONFERENCE ON GOVERNANCE
• Date of Conference: 19/12/2025 - 19/12/2025
• Venue: BARD Cumilla
• Organizer: BARD Cumilla