Klasifikasi Jenis Gonggong Melalui Pendekatan Pengenalan Objek Berbasis MobileNet-SSD

Abstract

Gonggong is a type of sea snail that is commonly consumed and has become an icon of the culinary specialties of the Riau Islands Province. The most commonly recognized and consumed types in the Riau Islands are Laevistrombus turturella and Strombus canarium. These two types of gonggong have similar physical characteristics and can be difficult to distinguish. Therefore, this research was conducted to find a practical solution that can classify types of gonggong based on their visual images. This study uses a real-time object detection approach based on the MobileNet SSD framework implemented in TensorFlow and applied to an Android-based mobile application. The dataset used consists of 418 images of both types of gonggong that have been augmented with and without backgrounds. The results of the tests show that the model has a confidence level of 83% for images without backgrounds, and 67% for images with backgrounds. These findings indicate that the method used has the potential for further development to improve the model's confidence level in classifying types of gonggong.