An Intelligent Waste-Sorting and Recycling Device Based on Improved EfficientNet

Abstract

The main source of urban waste is the daily life activities of residents, and the waste sorting of residents’ waste is important for promoting economic recycling, reducing labor costs, and protecting the environment. However, most residents are unable to make accurate judgments about the categories of household waste, which severely limits the efficiency of waste sorting. We have designed an intelligent waste bin that enables automatic waste sorting and recycling, avoiding the extensive knowledge required for waste sorting. To ensure that the waste-classification model is high accuracy and works in real time, GECM-EfficientNet is proposed based on EfficientNet by streamlining the mobile inverted bottleneck convolution (MBConv) module, introducing the efficient channel attention (ECA) module and coordinate attention (CA) module, and transfer learning. The accuracy of GECM-EfficientNet reaches 94.54% and 94.23% on the self-built household waste dataset and TrashNet dataset, with parameters of only 1.23 M. The time of one recognition on the intelligent waste bin is only 146 ms, which satisfies the real-time classification requirement. Our method improves the computational efficiency of the waste-classification model and simplifies the hardware requirements, which contributes to the residents’ waste classification based on intelligent devices.

Framework of Smart and Integrated Household Waste Management System: A Systematic Literature Review Using PRISMA

“…Based on The physical condition of the waste, the type of waste that dominates is solid waste [15,20,36,37,66]. Based on the risk level of waste, the most dominant type of hazardous waste [22,[40][41][42][43][44][45][46][47]. The selection of hazardous waste as a type of waste shows that there is a focus on efforts to manage hazardous waste so that it does not hurt the environment.…”

“…Digital cameras are used to take pictures of everyday waste [9,16,36], which can add to the collection of waste images [46]. Furthermore, waste images can be used for visual image analysis to be classified [43,44].…”

“…In the hardware subdimension, there is a need for sensor devices with functionality, namely to (1) detect the movement of people throwing rubbish into containers using motion sensors [51] and passive infrared (PIR) sensors [12]; (2) capturing images of trash using an image sensor [41]; (3) detecting the presence of waste using ultrasonic sensors [9,16,43] and infrared sensor [44]; (4) weighing the weight of the waste stored in the container [19,42] and to find out in more detail the type, weight and ratio of waste impurities [53] with using load-cell sensors [21,59]; (5) measuring the level of waste in the container using an ultrasonic sensor [21,22,37,49,59], overfill sensor [41], tracker sensor [43] infrared sensor [45]; (6) detecting the number of solid or liquid particles in the waste container using an ultrasonic sensor [12], (7) detecting wet or dry waste using a capacitive sensor based on water content [49], temperature and humidity sensors based on temperature and humidity levels [50], and moisture sensors and touch sensors [51]; (8) detecting the presence of metal waste using metallic sensors [49], induction sensors [46], and inductive proximity sensors [48,50]; (9) detecting the presence of plastic and wood waste using capacitance proximity sensors [48]; (10) storing and managing time in the waste recycling process, namely using a real-time clock (RTC) sensor [12].…”

“…Artificial intelligence technology to predict waste can use a multi-site Long Short-Term Memory (LSTM) neural network, which is an artificial neural network using a deep-learning approach [14], Random Forest and Knearest neighbor models [20], and ensemble learning [15]. Artificial intelligence technology that can be used to detect and classify waste includes machine vision technology and convolutional neural networks (CNN) algorithms [9,55], a neural network to train several photos of waste so that it can help in waste sorting based on color and location coordinates [36], multimodal cascaded convolutional neural network (MC-CNN) by combining DSSD, YOLOv4, and Faster-RCNN for automatic detection and classification of domestic waste [53], Lightweight CNN model using MobileNet V2 to accurately classify various types of waste, including recyclable waste, kitchen waste, hazardous waste and other waste [41], CNN and Random Forest to classify waste according to categories [43], CNN with the VGG16 type classifies captured waste images into categories of plastic, paper, and glass to order the trash container to open according to category [16], CNN with VGG16 identifies waste images with the registered dataset and sorts the waste [46], the GECM-EfficientNet model, which is part of AI, uses a dataset that has been trained to classify trash based on its categories [44], the K-nearest algorithm Neighbor (KNN) is used to classify bio and non-biodegradable waste levels and toxic gas concentrations [48]. The data recognition model can also be used for waste classification [40].…”

A Study on the Recycling Classification Behavior of Express Packaging Based on UTAUT under “Dual Carbon” Targets

“…Zhou, J et al [17] introduced a computer vision-based color-sorting system for waste textile recycling that can classify relevant colors well. Feng, ZC et al [18] designed an intelligent waste bin that can automatically classify and recycle waste, improving the efficiency of waste classification.…”