Smartphone-based bulky waste classification using convolutional neural networks
- PDF / 1,828,330 Bytes
- 21 Pages / 439.37 x 666.142 pts Page_size
- 97 Downloads / 269 Views
Smartphone-based bulky waste classification using convolutional neural networks Hanxiang Wang 1 & Yanfen Li 1 & L. Minh Dang 1 & Jaesung Ko 2 & Dongil Han 1 & Hyeonjoon Moon 1 Received: 5 November 2019 / Revised: 3 August 2020 / Accepted: 6 August 2020 # Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract
The rapid urbanization process is escalating the urban waste problem, and ineffective management has worsened the issue, leading to severe consequences to the population health and economy. Although many countries have started to charge money for large household items, it is time-consuming and challenging for collectors to distinguish various types of bulky waste manually. As a result, this study introduces a mobile-based automatic bulky waste classification system. The original contributions include (1) a fine-tuned VGG-19 model is proposed to classify 95 types of bulky wastes; (2) three hybrid models are introduced to efficiently handle the imbalanced data problem, including class-weight VGG-19 (CW_VGG19), eXtreme Gradient Boosting VGG-19 (XGB_VGG19), and Light Gradient Boosting Machine VGG19 (LGB_VGG19); (3) a large dataset that includes 95 classes, and each class contains over 500 images; and (4) the development of a mobile application that used the proposed model. Experiments show that the model obtained an accuracy of 86.19%, which outperforms existing models in classifying bulky waste. Moreover, the proposed hybrid models showed their robustness against imbalanced data under various scenarios. Keywords Waste classification . CNN . Imbalanced data . XGB . LGB . Bulky waste
1 Introduction Based on the United Nations statistics, about 60% of countries in the world emphasized their deep concern for managing waste and other environmental problems in the 2015 Earth Summit [23]. Therefore, it is crucial to reduce the environmental impact of waste through rational
* Hyeonjoon Moon [email protected]
1
Department of Computer Science and Engineering, Sejong University, Seoul, South Korea
2
Gatda corporation, 31 Baekbeom-ro 21-gil, Mapo-gu, Seoul, Republic of Korea
Multimedia Tools and Applications
waste sorting and management. According to statistics in 2015, South Korea is one of the world’s most densely populated nations with a density of 503 people per square kilometer [19, 26]. As a result, if ever-increasing garbage is not handled appropriately, severe environmental pollution is imminent, and the quality of life of the population will be significantly affected. For centuries, garbage disposal and management have been a thorny issue of great concern around the world. During the Middle Ages, in order to avoid the spread of diseases, governments took measures to remove garbage dumps as they were accelerating the increase of rat populations [1, 34]. The eighteenth century saw the start of recycling reusable materials, such as scrap metal, paper, and wood. Subsequently, modern industrial recycling began to appear in the nineteenth century [1, 24]. Through the end of the 1900s, with
Data Loading...
