Developed a computer vision system to detect crop pests in real time and support precision agriculture by enabling timely pest identification and targeted pesticide application. The project aimed to reduce excessive pesticide usage while helping improve crop health through early detection. I collected and prepared annotated image datasets, trained a YOLOv8 object detection model, and evaluated its performance across multiple pest classes. The system identifies pests directly from field images and estimates infestation severity, which can be used to recommend appropriate pesticide dosages based on the affected crop area. The solution demonstrates how deep learning can assist farmers in making faster and more informed pest management decisions. Achieved 99.5% mAP@50 for Bollworm detection, with 100% Precision and 98% Recall. Attained 96% mAP@50 for Stem Borer detection and 91% mAP@50 for Brown Planthopper detection. Successfully detected five different agricultural pest classes using a single YOLOv8 model. Implemented pest severity estimation to generate adaptive pesticide dosage recommendations based on crop area and infestation level. Built a complete pipeline covering dataset preparation, annotation, model training, evaluation, and real-time inference.