Machine Learning for Beach Litter Detection

摘要

People from economically weaker sections find consolation in doing unskilled tasks which are easily available, though a few of them are certainly not for humans. They have to be replaced with robots since these jobs fall into the categories of dull, dirty, difficult, and dangerous jobs. Exclusion of human involvement in the demeaning tasks and provision of hygienic environments around the beach areas would contribute to a better opportunity for a country's tourism. The task of implementation of beach cleaning with robots throws many technical challenges, a few of which are addressed in this research work. Machine learning has influenced the progress and outcomes for various domains of engineering and science including statistics. Even in social domains the impact of advent of Machine learning has been felt by not just the end users, but also the researchers. In this work, different methods for classifications of beach litter are proposed and evaluated. A dataset is collected and the classifiers are evaluated based on various metrics. An appropriate classifier is then selected based on these metrics, and the system is used on a beach cleaning robot.