Quantification of Litter in Cities Using a Smartphone Application and Citizen Science in Conjunction with Deep Learning-Based Image Processing

作     者：Kako, Shin’ichiro Muroya, Ryunosuke Matsuoka, Daisuke Isobe, Atsuhiko 

作者机构：Graduate School of Science and Engineering Department of Engineering Ocean Civil Engineering Program Kagoshima University Japan Japan Agency for Marine-Earth Science and Technology Japan Research Institute for Applied Mechanics Kyushu University Japan 

出 版 物：《SSRN》 

年 卷 期：2023年

核心收录：

主　　题：Object detection 

摘      要：We aimed to classify and quantify the abundance of litter in cities, which is a major source of marine plastic pollution, using citizen science combined with deep learning-based image processing. The project to collect image data including street litter through the smartphone application Pirika was launched in May 2018 and is ongoing and the total number of submitted images has reached one million. Of these, it was visually confirmed that images with clearly distinguished street litter accounted for approximately 30% of all images. Visual classification revealed that the most common types of litter were cans, followed by plastic bags, plastic bottles, cigarette butts, cigarette boxes, and masks. The top six categories account for approximately 80% of the total and the top three categories account for more than 60% of the total imaged litter. Based on these results, a classifier capable of classifying and quantifying these six litter types was constructed using image processing based on deep learning trained using the submitted images. This model was validated using images unused for training. The results revealed that both precision and recall derived from our model were higher than 75%. The quantity of litter derived from automated image processing was also plotted on a map using location data acquired concurrently with the images by the smartphone application. Therefore, we demonstrated that citizen science supported by smartphone applications and deep learning-based image processing is capable of visualizing/quantifying street litter in cities. © 2023, The Authors. All rights reserved.