Hybrid deep neural network identification coastal garbage to optimize the potential of sustainable marine resources in North Sumatera

Indonesia has around 8,090 coastal villages, and 60% of Indonesia’s total population lives in coastal areas. The majority of people living in coastal areas rely heavily on marine resources as their main livelihood. One of the dominant problems of coastal areas is the lack of optimal waste management solutions. The coastal waste problem is a serious challenge to the sustainability of Indonesia’s marine resources. Coastal debris can damage marine ecosystems, threaten marine life, and negatively impact the fisheries and tourism sectors. The research objectives are (1) analysis of hybrid deep neural networks, recurrent neural networks (RNN), feedforward neural networks for image generation, and identifying and classifying garbage in coastal areas. (2) Identifying garbage in the North Sumatra Coastal area with single-label, multi-label, and image recognition models. The stages of the research method carried out are (1) Literature study, (2) Problem identification and data collection of coastal waste, (3) Data preprocessing includes normalization, image resizing, and cleansing of waste data, (4) Analysis of the Hybrid Deep Neural Network (HDNN) Model, Testing in terms of accuracy, efficiency, and effectiveness. Next is (5) a Visual representation of waste accumulation in the coastal regions of North Sumatra and areas susceptible to pollution caused by garbage, and (6) a Model Evaluation. The results of the study found that the hybrid ANN method can be used to identify and classify coastal waste in North Sumatra based on image data with a precision value of 95.7%, F1 value of 95.2%, and Recall 95.2%. Furthermore, the results of HDNN image recognition are precision 63.2%, recall 92.3%, and F1 75%.