Implementasi Faster R-Cnn pada Klasifikasi Jenis Bunga Tanaman Salak Berbasis Desktop

Abstract

Indonesia, known as a tropical country, has a climate and soil that is fertile and rich. This advantage brings benefits with the many endemic and tropical plants that grow and produce fruit and vegetables with high nutrients and nutrition. As the healthy lifestyle trend grows and the purchasing power of local and foreign people for fruit increases, Indonesia has the opportunity to export its local fruits abroad. One of the well-known endemic tropical fruits in Indonesia is salacca. There are several types of salacca that grow in Indonesia, especially in North Sumatra, which are also popular with tourists, such as pondoh salacca, madu salacca, sidempuan salacca, and many more. The salacca plant is a dioecious plant, meaning that it has female flowers (putik) and male flowers (benang sari) on separate trees. This means that one salacca tree only has the status of a male or female tree, not both. Pollination of the salacca plant can be assisted by wind, but this affects the quality and quantity of the fruit that will grow later, so pollination will be more complete if assisted by humans. Not many people are aware of the differences between male and female flowers on the salacca plant, so the author is interested in using the salacca plant as an object of research in building a system for classifying objects based on images. The method applied is Faster R-CNN, which is known to have good performance in detecting objects. The research object focuses on 3 types of salacca flowers, namely pondoh female, pondoh male, and sidempuan male. The accuracy obtained as the result of the system's performance in detecting objects is 93%.