IMPLEMENTASI ALGORITMA CNN MOBILENET UNTUK KLASIFIKASI GAMBAR SAMPAH DI BANK SAMPAH

Abstract

Waste is a global problem that must be resolved so that the environment is well maintained, especially in developing countries such as Indonesia. A good understanding of waste will have an impact on optimal waste management. Waste banks are one way out in waste management so that what was originally useless becomes marketable. In its implementation, the waste bank must have data collection and transparency of customer waste data in real time. Technology is one of the main solutions to produce digital products that make it easier for people to access information accurately. By using the Convolutional Neural Network (CNN) algorithm, an image of image data can be predicted with good accuracy. Along with the development of technology today, various kinds of architectures are present, one of which is Mobilenet. This architecture has the ability to run machine learning models on mobile and IoT devices. Furthermore, the resulting model is quite good with an accuracy rate of 96% on Metal waste, 92% on Paper and Organic waste, 80% on Cardboard waste, 76% on Glass waste, and 72% on Plastic waste. The disadvantages that exist in this model when predicting with almost similar shapes and images that have many objects in them, the error will be greater so there is a possibility of error in predicting the results of the garbage image.