SINERGI Vol. 27, No. 2, June 2023: 163-170 http://publikasi.mercubuana.ac.id/index.php/sinergi http://doi.org/10.22441/sinergi.2023.2.003 K. Aziz et al., Multilabel image analysis on Polyethylene Terephthalate bottle images … 163 Multilabel image analysis on Polyethylene Terephthalate bottle images using PETNet Convolution Architecture Khoirul Aziz 1 , Inggis Kurnia Trisiawan 1 , Kadek Dwi Suyasmini 1 , Zendi Iklima 1 *, Mirna Yunita 2 1 Department of Electrical Engineering, Faculty of Engineering, Universitas Mercu Buana, Indonesia 2 Department of Computer Science, School of Computer Science and Technology, Beijing Institute of Technology, China Abstract Packaging is most of the important aspects of the product. Good packaging can increase the competitiveness of a product. Therefore, to maintain the quality control of the packaging of a product, it is necessary to have a visual inspection. Furthermore, an automatic visual inspection can reduce the occurrence of human errors in the manual inspection process. This research will use the convolution network to detect and classify PET (Polyethylene Terephthalate) bottles. The Convolutional Neural Network (CNN) method is one approach that can be used to detect and classify PET bottle packaging. This research was conducted by comparing seven transfer learning models of CNN, namely VGG-16, Inception V3, MobileNet V2, Xception, Inception ResNet V2, Depthwise Separable Convolution (DSC), and PETNet, which is the architectural model proposed in this study. The results of this study indicate that the PETNet model gives the best results compared to other models, with a test score of 96.04%, by detecting and classifying 461 of 480 images with an average test time of 0.0016 seconds. This is an open access article under the CC BY-SA license Keywords: Convolutional Neural Network (CNN); Polyethylene Terephthalate Network (PETNet); Quality Control; Transfer Learning Model; Visual Inspection; Article History: Received: April 9, 2022 Revised: October 10, 2022 Accepted: October 16, 2022 Published: June 2, 2023 Corresponding Author: Zendi Iklima Electrical Engineering Department, Universitas Mercu Buana, Indonesia Email: zendi.iklima@mercubuana.ac.id INTRODUCTION In the product packaging process that uses bottles as containers for the product, several stages must be passed before it becomes a finished product (a finished good), including filling, labeling, capping, and others [1][2]. Of course, these stages cannot be separated from errors in the product packaging process [3]. Moreover, when defective products are distributed to consumers, they may generate complaints or even product returns [4][5]. Therefore, manufacturers must use a quality control system to reduce financial risk and reputational damage [6]. Thorough testing of products is required before shipment [7]. As a result, in order to maintain production quality, a visual inspection of the product is required to ensure that the results are in accordance with the established rules. One of the most common procedures in the industry is the visual approach to defect detection [8]. Most bottle inspections are hand-picked by operators using eye checks [9]. But currently, due to increased production capacity, an industry cannot use human labor to sort a product [10]. Therefore, many industries have recently incorporated artificial intelligence (AI) algorithms into their manufacturing processes [11][12]. In particular, AI approaches within the industry are always needed to reduce machine failures and improve quality control products automatically [13]. Intelligent production planning systems will increase industry efficiency and productivity [14]. One of the numerous digital image classification methods available today that can be used for visual sorting is the Artificial Neural Network (ANN) method. An ANN is a traditional structure comprising three layers: input, output, and hidden. There are a different number of neuron elements in each layer [15]. Multi-Layer Perceptron (MLP) is the name given to this type of ANN model with many layers (MLP). In terms of classification, MLP is extremely accurate. MLP, on the other hand, has a weakness in digital image classification. To address this problem, MLP, specifically the Convolutional Neural Network