Nowadays, digital technology and automation are widely applied in agriculture to evaluate field trials. Early detection of pests and diseases on the plants is crucial, as it significantly impacts the economy and relates to essential human food sources. However, early detection is a significant challenge for farmers, as pests are often tiny and have colors similar to the flowers, fruits, and leaves they damage. In addition, pests often hide and move quickly in the natural environment, making detection and treatment often occur late, leading to severe consequences. In this study, we tested the YOLOv9c model to detect pests in agriculture. We compared YOLOv9c with other versions such as YOLOv8n, YOLOv6n, YOLO6m, YOLOv5n, and other methods to gain an overview. In order to improve the performance of the model, we adjusted the hyperparameters during the training process, such as epochs, batch, LR0, and weight decay. Thus, YOLOv9c has demonstrated high effectiveness, with a mAP@5 result of 95.2%, an F1-score of 92.7%, precision of 92.6%, and recall of 92.9%. Our contributions indicate that YOLOv9c outperforms other YOLO versions. Based on the experiments, we have implemented innovative solutions such as Programmable Gradient Information (PGI) and Generalized Effective Layer Network (GELAN), setting a new precedent for future research and applications in this field.