Deteksi Kematangan Cabai Rawit untuk Benih dengan Menggunakan SSD-Mobilenet secara Realtime

Abstract

The cultivation of bird's eye chili is highly sought after by the Indonesian community. However, despite this popularity, many Indonesian farmers face difficulties in cultivating bird's eye chili, with one of the primary challenges being the seed factor. Farmers encounter challenges related to seed issues in bird's eye chili cultivation due to the high cost of seeds and the uncertainty regarding their quality. Consequently, farmers need the ability to produce their own seeds. Farmers are required to select high-quality fruits to use as seeds, and the process of fruit selection for seed purposes is conventional and time-consuming, especially for those new to bird's eye chili cultivation. With the advancement of technology in the fields of Computer Vision and Artificial Intelligence, these tools are leveraged for fruit detection, making it feasible on mobile devices. The objective of this research is to develop a system capable of real-time classification of bird's eye chili fruits based on their ripeness levels. This is achieved by implementing SSD-MobileNet as the neural network architecture. The system is designed to be mobile-based. The research employs a dataset consisting of 1210 entries, with 880 used as training data and 220 as validation data. Additionally, 110 real-time data points are gathered for testing using a smartphone camera. The application of the SSD-MobileNet method in classifying the ripeness levels of bird's eye chili fruits resulted in an accuracy of 93.6%. This outcome indicates that the system, developed using the SSD-MobileNet method, has successfully and effectively classified the ripeness levels of bird's eye chili fruits for seed purposes.