Abstract
West Nusa Tenggara province has an average population engaged in farming. Farmers in West Nusa Tenggara usually cultivate rice crops in all of their paddy fields, as rice is a staple food for the Indonesian community. Therefore, rice production needs to increase every year, while rice harvests in West Nusa Tenggara have experienced fluctuations in the past three years, despite the requirement for high production. These fluctuations are caused by bird pests in rice crops, resulting in a loss of harvest yield of up to 30-50%. To address this issue, this research aims to develop an artificial intelligence model to detect bird pests in rice crops in real-time using the YOLOv5 method, specifically the S type. The YOLOv5 method was chosen for its ability to detect objects with high accuracy and fast processing speed, as demonstrated in previous research in other domains. However, there has been no research implementing this method specifically for detecting bird pests in rice crops. In this study, the authors developed a YOLOv5s model trained with 100 epochs and a batch size of 16. The training results showed the best accuracy of 96%. This finding indicates that the YOLOv5s model has great potential in detecting objects resembling bird pests in rice crops. The successful development of this model contributes significantly, both scientifically and practically, to efforts in reducing harvest losses caused by bird pest attacks. In further research, it is hoped that the development of this model can be expanded to detect other types of bird pests and widely applied in agricultural practices in West Nusa Tenggara and other regions.
Concepts :
Access to Document
10.1109/icadeis58666.2023.10270848SDGs
Citations by Year
| Year | Count |
|---|---|
| 2023 | 4 |