Deteksi Tingkat Kematangan Buah Kopi Menggunakan SSD – Mobilenet secara Realtime

Abstract

Coffee plants are one of the plantation crops with distinctive flavors and aromas that have high economic value and are commercially developed. The best coffee beans come from coffee cherries that have met the harvesting standards, namely perfectly ripe coffee cherries. The process of harvesting coffee fruit in Indonesia generally still applies traditional methods manually by paying attention to the characteristics of the harvested coffee fruit. If concluded based on the physical characteristics of the coffee fruit, each farmer has a different assessment of the color and shape of the coffee fruit. With the advancement of technology in the field of Computer Vision and Artificial Intelligence, a mobile-based system will be built that is able to facilitate coffee farmers in detecting the level of maturity of coffee fruit automatically by only taking images of coffee fruit in real time. This research will utilize SSD-Mobilenet as its network architecture. The amount of training data used in this study was 3,180 data and the testing data was 636 data. The detected coffee fruit ripeness level is divided into 4 classes, namely raw, half-ripe, ripe and perfectly ripe. The test results show that the system developed is able to detect the maturity level of coffee fruit with an accuracy rate of 94.34%.