SISTEM PENDETEKSI SAMPAH BOTOL PLASTIK DI BAWAH LAUT MENGGUNAKAN METODE DEEP LEARNING SECARA REAL TIME = Underwater Plastic Bottle Trash Detection System Using Real Time Deep Learning Method

Based on data from a study titled "Plastic Waste Discharges from Rivers and Coastlines in Indonesia," Indonesia is estimated to generate around 7.8 million tons of plastic waste every year. Plastic waste is a major source of marine debris worldwide and presents serious challenges in the decomposition process in the aquatic environment. All kinds of plastic and non-plastic waste originate in river waters and end up in marine and coastal ecosystems. As a result, when plastic waste settles in marine waters, it will have a negative impact on the reproduction and migration process of marine animals, inhibit coral growth, damage marine habitats, and trigger death in marine animals that interact with or are trapped in plastic waste. This research develops a real time underwater plastic bottle waste detection system using YOLOv8 deep learning algorithm. The dataset in this study is primary obtained from taking pictures of plastic bottle waste directly in the sea and in swimming pools. The results show that the model built has a precision level that reaches 99.6% and recall reaches 100%. The results of the detection performance evaluation using the Intersection over Union (IoU) threshold of 50% showed that the mean Average Precision at 50 (mAP50) reached 99.5%, confirming the quality of the model in measuring detection accuracy at that level. In fact, the mAP50-95 value reached 97.8%, indicating the consistency of the model's performance in detection over the IoU range from 50 to 95%. This research successfully developed a real time underwater plastic bottle waste detection model using the YOLOv8 approach. Thus, this research opens the potential for further development to reduce the negative impact of plastic waste on aquatic ecosystems in Indonesia.