Implementasi Deteksi Objek Real-Time Sebagai Media Edukasi dengan Algoritma YOLOv8 pada Objek Sampah

Waste is one of the complex global issues and is one of the points of the SDGs indicators related to municipal waste, food waste, hazardous waste and recycling systems that cannot be resolved. According to data from SIPSN in 2022, waste generation in Indonesia will reach 35,289,535.55 tons/year while about 47.32% of waste handling is done, which is 16,697,790.76 tons/year. The National Research and Innovation Agency said that currently there are still few Indonesians who have the awareness to start sorting waste from their own homes. As many as 80% of Indonesians do not sort their waste. To overcome these problems, everyone needs to make changes early on by making waste management a habit so as to change people's skeptical attitude towards waste management. Thus, research was conducted to identify the type of waste using the YOLOv8 algorithm, with a dataset of 17,617 data which was then analyzed by creating a yolov8 model. The best accuracy results were obtained by using the yolov8l variant as well as 16 batch sizes and the SGD optimizer with a learning rate value of 0.001 as a parameter. The model training process was then evaluated using confusion matrix with a percentage reaching 86.5%.