Conference PaperPDF Available

Abstract and Figures

ResearchAndMarkets wrote in their report on May 15, 2018, that up to 1.2 Trillion USD in 2017 of products are counterfeited goods. The report estimated this damage globally at 1.82 Trillion USD in 2020 (RESEARCH AND MARKETS, 2018). This paper does not consider copyrights or digital piracy, counterfeiting, fraudulent documents but rather investigates the prevention of counterfeiting on a technological basis. The presence of counterfeit products on the European market is on the increase, therefore the intervention of inspection bodies and authorities alone is not sufficient, consumers can make their contribution and support this process. In this paper, we research the possibility to reduce counterfeit products using machine learning-based technology. Image and text recognition and classification based on machine learning have the potential to be a key technology in the fight against counterfeiting. The automatic image and text recognition and the classification of product information enable end customers to detect counterfeits precisely and quickly by checking them against trained models. The goal of this paper is to create an easy to use applications in which the end-user identifies the counterfeit product and contribute to the fight against product piracy.
Content may be subject to copyright.
IMPROVING FAKE PRODUCT DETECTION
USING AI-BASED TECHNOLOGY
Eduard Daoud, Dang Vu, Hung Nguyen and Martin Gaedke
Technische Universität Chemnitz, Germany
ABSTRACT
ResearchAndMarkets wrote in their report on May 15, 2018, that up to 1.2 Trillion USD in 2017 of products are
counterfeited goods. The report estimated this damage globally at 1.82 Trillion USD in 2020 (RESEARCH AND
MARKETS, 2018). This paper does not consider copyrights or digital piracy, counterfeiting, fraudulent documents but
rather investigates the prevention of counterfeiting on a technological basis. The presence of counterfeit products on the
European market is on the increase, therefore the intervention of inspection bodies and authorities alone is not sufficient,
consumers can make their contribution and support this process. In this paper, we research the possibility to reduce
counterfeit products using machine learning-based technology. Image and text recognition and classification based on
machine learning have the potential to be a key technology in the fight against counterfeiting. The automatic image and
text recognition and the classification of product information enable end customers to detect counterfeits precisely and
quickly by checking them against trained models. The goal of this paper is to create an easy to use applications in which
the end-user identifies the counterfeit product and contribute to the fight against product piracy.
KEYWORDS
Anti-Counterfeiting, Machine Learning, Deep Learning, Image Recognition, Object Detection
1. INTRODUCTION AND CURRENT PROBLEM
Detection of counterfeit products is in certain cases a challenge for the consumers and can sometimes be
dangerous when it comes to medical products or toys for children, for example. ResearchAndMarkets wrote in
their report on May 15, 2018, that up to 1.2 Trillion USD in 2017 of products are counterfeited goods. The
report estimated this damage globally at 1.82 Trillion USD in 2020 (RESEARCH AND MARKETS, 2018).
Especially The presence of counterfeit products on the European market is on the increase (OECD/EUIPO,
2016), therefore the intervention of inspection bodies and authorities alone is not sufficient. Since its launch in
2003, the EU's Rapid Alert System has been providing EU member States with a network and communication
tools to publicize counterfeit products. The system stabilizes at regular intervals, about 50 alerts are published
each week on the European Commission's website, with slightly more than 2,000 alerts released each year.
(Directorate-General for Justice and Consumers (European Commission), 2018). The number of counterfeits
reported products is extremely low in relation to the number of counterfeit products imported into the EU.
OECD wrote in their report 2016 that up to 5% of imports are counterfeited goods. The report estimated this
damage at EUR 85 billion (OECD/EUIPO, 2016). A major problem of such governmental instruments is that
the end-consumers is not involved, if at all, in the detection process of counterfeiting.
In the next section, we will highlight the subject of counterfeit domains and focus on the area where the use
of IT technology can make a positive contribution. After introducing the related works, we will outline the
solution concept and technical architecture, then we will focus on the implementation and evaluation of such
solutions and their challenges. Finally, we will review the results of our work and consider the outlook for the
future.
18th International Conference e-Society 2020
119
2. RELATED WORKS
In general, anti-counterfeiting technologies should be easily applied and used by end consumer first but
challenging to imitate and ensure that businesses are safeguarded financially and in terms of reputation. Based
on (Li, 2013) these technologies usually have four main features:
difficult to duplicate or forge,
easily identifiable visually without the need of special equipment,
hard to re-label or reuse, and
easily noticeable when tampered with.
From a product standpoint, there are three common categories for anti-counterfeiting: overt, covert, and
track and trace, shown in Figure 1. Overt technologies focused on the packaging of the products.
Color-shifting inks, watermarks and holograms are some of the technologies that can be used in this category.
End consumers need to be briefed in advance so that they interpret these technologies correctly to verify the
fake products. Covert technologies like ultraviolet (electromagnetic radiation) and bi-fluorescent are also
applied to the product itself but are not identifiable without special equipment. Digital watermarks, hidden
printed messages and pen-reactive ink. are also part of the covert technologies.
The final category is track and trace. Radio Frequency Identification (RFID) tags, Electronic Product
Codes (EPCs) and barcodes are the main technologies in this category. The effect of holistic tracking and
tracing contributes that the reduction of counterfeits is possible. Consumers and retailers scan the code already
implemented by suppliers and manufacturers to verify the authenticity of the product or to trace the overall
supply chain process.
Figure 1. Anti-counterfeiting technologies
ISBN: 978-989-8704-14-6 © 2020
120
There are other approaches based on improved communication between companies and organizations with
the interest to reduce counterfeiting on the market - an example is React. React is a not-for-profit organization
providing a market, online and customs enforcement professional services (React, 2020). Professional services
approaches have a big advantage concerning accuracy but still working with the manual process and need
manpower. All three technologies mentioned in Figure 1. Anti-counterfeiting technologies have disadvantages
and limitations. In previous researches, we have addressed the subject in detail but with the use of blockchain
technologies (Daoud & Gaedke, 2019) and we found also a lot of limitations. Counterfeiters are becoming
more and more professional and sophisticated. They are always developing approaches to better package
counterfeit products and bring them to the market undetected (Shields & Deshmukh, 2020). Consequently,
overt and covert technologies can be easily imitated. It is difficult for the average end-consumer can
distinguish between a convincing imitation and the real product. Besides, covert technologies are required
special devices to identify counterfeited products so that customers are neither able to detect nor verify their
products. The track and trace technologies with its encoding and security feature can be used in combination
to improve this situation but it leads to another major issue: overhead cost. As a consequence, the price of the
product has risen further, which in turn leads the end-user to look for counterfeit products in certain cases.
(Li, 2013).
Based on that, this paper proposes a low-cost and user-friendly solution relies on machine learning-based
technology which enables end-consumers to identify and verify products without any special equipment. By
using image and text recognition, this approach aims to improve fake product detection. It can also combine
with over, covert and/or Track and Trace technologies to help combat counterfeiting more efficient and
effective. In the next section, we will introduce the concept behind the solution and the technical architecture
after that we will go deeper in the implementation stage.
3. CONCEPT AND TECHNICAL ARCHITECTURE
According to the report of (Statista, 2019), the current number of mobile phone users in the world is
4.78 billion, of which 3.5 billion are smartphone users. Today, with cheap price, users can easily own a
smartphone with a built-in digital camera and internet access. Based on that, the proposed solution will allow
the end-consumers to use their phones as equipment to detect fake products.
For detection, the end-consumer takes pictures of a product packaging, which contains product text
information, logos and maybe certification marks/logos. These pictures will be sent in a request to the server
for processing and verifying. Afterwards, the detection result will be returned to the end-consumer to make a
further decision. In the case of fake product detection, the end-consumer has an ability to report this counterfeit
product to the government system, such as the Safety Gate - EU's Rapid Alert System. An overall concept of
this solution is shown in Figure 2.
Figure 2. Overall Concept
18th International Conference e-Society 2020
121
In the server-side, there are two main components: a web server and a machine learning application. The
web server will be acted as the middle layer. It receives requests (pictures and metadata) from the end-users
(the client-side) and forwards this request information to the machine learning application for detection. After
processing, the web server receives the detection result and sent the response back to the client. In addition,
the server also performs several operations such as storing detection results, data statistic or allowing users to
report counterfeit products.
The machine learning application is the main contribution of this paper. An architecture of this application
has three main parts: the dataset, detection models, and trained model. In order to have an efficient machine
learning application, we need first a dataset represents the related entities and their values. There are two types
of dataset: training dataset and test dataset. In our application, dataset is a collection of valid logo or/and mark.
Next, we need the detection models- the functions/algorithms which perform certain operations on the given
input (product images with mark/logo) and procedure the suitable output (detection result). More specifically,
we apply the pre-trained Deep Learning (DL) detection models, e.g. SSD (Liu, et al., 2016) and Faster R-CNN
(Ren, et al., 2016), which will be discussed in detail in section 4. To create the trained model, we do a transfer
learning by training the pre-trained detection models with our input dataset. Our trained model contains two
key components: image detection and text recognition, which aims to verify the validity of the mark/logo.
Image detection and text recognition can be applied in a wide range of application, from optical character
recognition - OCR (machine translation, (extracted) text-to-speech), face detection, face recognition,
self-driving cars, to medical imaging, and robotics (Vahab, et al., 2019).
To challenge the concept and the proposed solution is the first machine learning application particularly
concentrating on detecting fake product, we implement the solution in the next section and training the model
with the certification body logo and the mark of TÜV SÜD AG. There are many reasons to assume, and based
on the (OECD/EUIPO, 2016), that many counterfeiters use the certification agency logo to gain better access
to the EU market.
4. IMPLEMENTATION
Figure 3 shows a detailed implementation of the approached machine learning detection solution. The
implementation consists of two steps: training models step and detecting logo step. In the first step, the training
model component depicts a way we prepare dataset and select the Deep Learning pre-trained detection model.
For demonstration, we collect 2000 certification marks and logos from the websites (web crawler) of the
Testing, Inspection and Certification (TIC) members, e.g. TÜV SÜD AG, Dekra and Bureau Veritas. To make
our dataset more diverse and avoid overfitting, we use another technique also to enrich our dataset by extracting
the certification marks and logos from video frames.
The more data we have, the higher the accuracy of our model can perform. After collecting dataset, we
label our raw data by using annotation tools (i.e., LabelImg) and divide them into a training set (80%) and test
set (20%). The collecting-labelling is not a one-time process. We operate it in an iterative fashion so that the
machine learning model could gain the ability to generalize and return more precisely outcome.
ISBN: 978-989-8704-14-6 © 2020
122
Figure 3. Implementation of machine learning detection application
Instead of investing time and effort for building a trained model from scratch, we use transfer learning- an
optimization that allows rapid progress and boots performance when modelling our detection task. Two
pre-trained models are used in our transfer learning are SSD300 - Regression/Classification based family- and
Faster R-CNN Resnet101 Region-based family-. Based on (Zhao, et al., 2019), Faster RCNN archive high
accuracy (mean Average precision of 83.8% by testing on VOC 07 Test set) with low training speed.
Because of efficiently detecting small image resolution with detailed features, it is used to verify
certification marks. On the other hand, SSD can run fast and smooth (test time of 0.02 second/image and rate
of 45 frames/second, tested on VOC 07 Test set) on identifying large objects, such as logo, with high precision.
The second step focuses on handling request from the webserver (product images with mark/logo) and
returns detection result. When an image is passed through our trained model, a prediction map will be returned
to classify and localize the certification mark/logo we need to detect. To perform visualizations for end-user,
the prediction bounding box and the class label for each certificate mark are drawn on the uploaded, shown in
Figure 4 in the next section. For demonstration, we construct a web application by using Flask - a lightweight
Python web application framework and Tensorflow - an end-to-end open-source library to deploy our machine
learning application. In the following section, we discuss the challenges of our approach bevor we headline the
summary and an outlook on our work.
5. EVALUATION AND CHALLENGES
The approached solution has several improvements for the current anti-counterfeiting technologies. Firstly, it
can be verified by average end-users, which can add a new protection layer to combat counterfeiting products.
In comparison to the track and trace technologies mentioned in section 2, our solution provides a low-cost
implementation, which is appropriate when the market is scaling up.
18th International Conference e-Society 2020
123
In addition, unlike overt technology, end-consumer does not need any special device to use our solution
they only need a mobile application and internet access to verify the genuineness of certification logo/marks.
The detection result from our demonstration reveals the potential of machine learning-based technology to
fight against counterfeiting. Our solution archives 97% precision at 3.1 seconds/certificate mark, on 400 tested
data. As the result in Figure 4, the sophisticated forged marks can be detected, e.g. minor change in color,
missing or incorrect text.
Figure 4. Tested results on TÜV SÜD AG certification mark
Similar to any other anti-counterfeiting technologies, issues and challenges arise for our machine
learning-based approach. One of the challenges is the limitation of data for the training process. The amount
of needed data to combat counterfeiting products is huge, but the existing data sources are limited. Furthermore,
crawling data from the Internet and manually annotating consumes a lot of time. The more data we have, the
more accuracy we can provide. Last but not least, 3.1 seconds for single certification mark detection is not
efficient in real-time application. Hence, there is still room for improvement, e.g. considering other faster
algorithms like YOLO (Redmon & Farhadi, 2019) but with the trade-off of accuracy. After highlighting the
challenges, we headline a summary and an outlook on our work.
6. CONCLUSION AND OUTLOOK
This paper presents a new approach as an anti-counterfeiting machine learning-based solution to detect fake
product. We evaluate the solution by using machine learning and found a lot of limitations. The machine
learning-based approach used in core deep learning and neural network technologies. The conclusions we can
define from the new approach are that the implementation of the system should be deeper researched, from the
point of view of collecting more training data and annotation/labelling service. The main focus is on which
implementations have a positive impact on anti-counterfeiting of products and the adoption of machine
learning-based detection depends on how the consumer can easily and simply interact with the system. By
using image recognition, this approach could improve fake product detection. It can also combine with over,
covert and/or track and trace technologies to help combat counterfeiting more efficient and effective.
In future work, we plan from one hand to explore and research more about using faster machine learning
algorithms to classify marks and logos and detect text with the help of OCR. from the other hand, we need to
extend our web crawler to have the possibility to gather more web information, especially from the eCommerce
world to find fake products with help of detecting logo, marks and text. This would combine three
state-of-the-art technologies, machine learning, Text recognition and web searching in one application. That
will bring great convenience and a better experience for users. However, we trust that using machine learning-
based technology will change the role and empower the consumer to drive the market for more transparency
and safety.
ISBN: 978-989-8704-14-6 © 2020
124
REFERENCES
Daoud, E. & Gaedke, M., 2019. Decentralizing Products Certificates Using Blockchain. Cagliari, Italy, 18th International
Conference on WWW/Internet, p. 8.
Directorate-General for Justice and Consumers (European Commission), 2018. Results of the EU rapid alert system for
dangerous non-food products, Brussels: European Commission.
Li, L., 2013. Technology designed to combat fakes in the global supply chain. Business Horizons, 56(2), pp. 167-177.
Liu, W. et al., 2016. SSD: Single Shot MultiBox Detector. s.l., arXiv preprint, arXiv:1512.02325.
OECD/EUIPO, 2016. Trade in Counterfeit and Pirated Goods: Mapping the Economic Impact. Paris: OECD Publishing.
React, 2020. [Online] Available at: https://www.react.org/ [Accessed 07 02 2020].
Redmon, J. & Farhadi, A., 2019. Yolov3: An incremental improvement. s.l., arXiv preprint arXiv:1804.02767.
Ren, S., He, K., Girshick, R. & Sun, J., 2016. Faster R-CNN: Towards Real-Time Object Detection with Region Proposal
Networks. s.l., In Advances in neural information processing systems (pp. 91-99)..
RESEARCH AND MARKETS, 2018. Global Brand Counterfeiting Report, s.l.: RESEARCH AND MARKERTS.
Shields, D. & Deshmukh, R., 2020. Anti-Counterfeit Packaging Market - Food And Pharmaceuticals by Authentication
Packaging Technology., 97220 United States: Global Opportunity Analysis and Industry Forecast.
Statista, 2019. Number of smartphone users worldwide from 2016 to 2021, s.l.: Statista.
Vahab, A., Naik, M. S., Raikar, P. G. & R, P. S., 2019. Applications of Object Detection System. International Research
Journal of Engineering and Technology (IRJET), 06(04), pp. 4186-4192.
Zhao, Z., Zheng, P., Xu, S. & Wu, X., 2019. Object detection with deep learning: A review.. EEE transactions on neural
networks and learning systems, 30(11), pp. 3212-3232.
18th International Conference e-Society 2020
125
... Counterfeited products have been a huge problem to the whole world market for a long time, there have been many types of research focused on solving this specific problem and only one solution was found proposing the use of machine learning and object detection to allow end-user to be able to verify and authorize products using only an image that contains the product quality certificate logos and marks. This solution was published by Eduard, et al. (2020) and it uses SSD by Liu, et al. (2016) and Faster R-CNN by Ren, et al. (2016) machine learning algorithms for detecting logos and marks in images. Logos and marks in product images are likely to be small, and these images will not always be in a high resolution, and that is one of the difficulties that make it hard for the SSD algorithm to detect and verify products efficiently. ...
... The two-stage detector with a focus on Faster R-CNN implementation by Girshick (2015) and the single-stage detector with a focus on SSD and YOLOv4 implementation by Alexey & Chien-Yao (2020). The CertificateOk platform resulting from the research of Eduard, et al. (2020) uses the Faster R-CNN implementation and the proposed new implementation uses the YOLOv4 implementation as an object detection algorithm. Figure 2 highlights the difference between the internal structure of one-stage detectors and two-stage detectors. ...
... We are going to use the dataset that was used in the previous implementation by Eduard, et al. (2020) of CertificateOk 1 and improve it a bit by adding more samples, the dataset that was used contained around 600 images of different shapes and different logo/marks dimensions. ...
Conference Paper
Full-text available
Counterfeit products are a major problem that the market has faced for a long time. According to the Global Brand Counterfeiting Report 2018, "Amount of Total Counterfeiting, globally has reached to 1.2 Trillion USD in 2017 and is Bound to Reach 1.82 Trillion USD by the Year 2020" a solution to this concern has already been researched and published by the authors in previous research papers published in e-society 2020 and IADIS journal. However, the issue with the previously mentioned solution was that the object detection performance and accuracy in detecting small objects need to be improved. In this paper, a comparison between the current state of the art algorithm YOLO (You Only Look Once) used in the new implementation and the SSD (Single Shot Detector) algorithm, the faster R-CNN (Region-Based Convolutional Neural Networks) used in the old implementation is made under the same condition and using the same training, testing, and validation sets. The comparison is made in the context of the present task to discuss and prove why YOLO is a more suitable option for the counterfeit product detection task.
... Counterfeited products have been a huge problem to the whole world market for a long time, there have been many types of research focused on solving this specific problem and only one solution was found proposing the use of machine learning and object detection to allow end-user to be able to verify and authorize products using only an image that contains the product quality certificate logos and marks. This solution was published by Eduard, et al. (2020) and it uses SSD by Liu, et al. (2016) and Faster R-CNN by Ren, et al. (2016) machine learning algorithms for detecting logos and marks in images. Logos and marks in product images are likely to be small, and these images will not always be in a high resolution, and that is one of the difficulties that make it hard for the SSD algorithm to detect small objects efficiently. ...
... The two-stage detector focuses on Faster R-CNN implementation by Girshick (2015) and the single-stage detector focuses on SSD and YOLOv4 implementation by Alexey & Chien-Yao (2020). The CertificateOk platform resulting from the research of Eduard, et al. (2020) uses the Faster R-CNN implementation and the proposed new implementation uses the YOLOv4 implementation as a framework. As in Manuel, et al. (2020) & Shilpa, et al. (2021), the Faster R-CNN contains two main elements: The region proposal network (RPN), and the Fast R-CNN header network. ...
... We are going to use the dataset that was used in the previous implementation by Eduard, et al. (2020) of CertificateOk 1 and improve it a bit by adding more samples, the dataset that was used contained around 600 images of different shapes and different logo/marks dimensions. After creating the dataset, a suitable labeling tool was used like YOLO_Label2, which is an open-source labeling tool that creates label files easily and efficiently. ...
Conference Paper
Full-text available
Counterfeit products are a major problem that the market has been facing for a long time. According to the Global Brand Counterfeiting Report 2018 "Amount of Total Counterfeiting, globally has reached to 1.2 Trillion USD in 2017 and is Bound to Reach 1.82 Trillion USD by the Year 2020", a solution to this concern has already been researched and published by the authors in previous research papers published in e-society 2020 and IADIS journal, but the issue with the previously mentioned solution was that the object detection performance and accuracy needed to be improved. In this paper, a comparison between the current YOLO (You Only Look Once) algorithm used in the new implementation and the SSD (Single Shot Detector) algorithm, the faster R-CNN (Region-Based Convolutional Neural Networks) used in the old implementation, is made in the context of the present task to discuss and prove why YOLO is a more suitable option for the counterfeit product detection task.
... Training models step and detecting logo step are the two steps required. Faster R-CNN achieves high accuracy and low training speed [13]. Chen and Shi et.al explains SCQI. ...
... In the RFID based system that low Cost RFID tags can be used for auto identification of products, but due to cloning of RFID tags, this method is not suitable [14]. In AI and machine learning application, CNN takes more time and memory.It needs training and testing phase before its actual deployment.Artificial Intelligence fails to detect tag reapplication attacks, wherein a counterfeiter removes a legitimate tag from a genuine product and reapplies it to a counterfeit or expired product [13]. There is no power for the customer,suppliers and retailers to check the integrity of product. ...
Article
Full-text available
Supply chain management frequently faced issues such as service redundancy, poor coordination between several departments, and lack of standardization as a result of the lack of transparency. Product counterfeiting is something which is very common now-a-days and it’s almost impossible to detect a counterfeit product just by looking at it. Counterfeiters cause significant challenges for legitimate firms, yet far too many people have no idea of the entire amount of counterfeit items’ influence on brands. There are several methods devised in the past to get away with this problem of product counterfeiting. The most popular methods are using RFID tags, Artificial Intelligence, QR code based systems, etc. But each of them had few disadvantages such as the QR code can be copied from a genuine product and placed on a fake product, artificial intelligence uses CNN and machine learning which needs heavy computational power and so on. The idea of this project is to improve detection of fake products by tracking its supply chain history. This is achieved with Blockchain technology which ensures the identification and traceability of real products throughout the supply chain. Blockchain based system, makes everything decentralized that may be accessed by several parties at the same time. One of its main advantages is that the recorded data is difficult to change without the consent of all parties concerned which makes the data extremely secure and protect from all vulnerabilities. This paper presents system designed using blockchain technology for detection of counterfeit products.
... Organisation of this paper-Section 1 explains the introductory area of AI and COVID-19, Section 2 depicts the relevant efforts proposed in these areas, Section 3 represents the technique for establishing one such bot, Section 4 demonstrates performance assessment and concludes with the details in Section 5. Eduard et al. (2020) [14] investigated the use of machine learning-based technologies for eliminating counterfeit items. Machine learning-based image and text identification and categorisation can be a crucial tool in the battle against counterfeiting. ...
Article
Full-text available
As the world is now fighting against rampant virus COVID‐19, the development of vaccines on a large scale and making it reach millions of people to be immunised has become quintessential. So far 40.9% of the world got vaccinated. Still, there are more to get vaccinated. Those who got vaccinated have the chance of getting the vaccine certificate as proof to move, work, etc., based on their daily requirements. But others create their own forged vaccine certificate using advanced software and digital tools which will create complex problems where we cannot distinguish between real and fake vaccine certificates. Also, it will create immense pressure on the government and as well as healthcare workers as they have been trying to save people from day 1, but parallelly people who have fake vaccine certificates roam around even if they are COVID/Non‐COVID patients. So, to avoid this huge problem, this paper focuses on detecting fake vaccine certificates using a bot powered by Artificial Intelligence and neurologically powered by Deep Learning in which the following are the stages: a) Data Collection, b) Preprocessing to remove noise from the data, and convert to grayscale and normalised, c) Error level analysis, d) Texture‐based feature extraction for extracting logo, symbol and for the signature we extract Crest‐Trough parameter, and e) Classification using DenseNet201 and thereby giving the results as fake/real certificate. The evaluation of the model is taken over performance measures like accuracy, specificity, sensitivity, detection rate, recall, f1‐score, and computation time over state‐of‐art models such as SVM, RNN, VGG16, Alexnet, and CNN in which the proposed model (D201‐LBP) outperforms with an accuracy of 0.94.
Conference Paper
Full-text available
OECD wrote in their report 2016 that up to 5% of imports are counterfeited goods. The report estimated this damage at EUR 85 billion (OECD/EUIPO, 2016). The presence of counterfeit products on the European market is on the increase, therefore the intervention of inspection bodies and authorities alone is not sufficient, consumers can make their contribution and support this process. 2003 started the EU a program “Rapid Alert System” as an information platform to publish alerts about dangerous products in the market. Over 2 000 alerts published by the system yearly. Thus, allowing local authorities to screen their market and take appropriate action should the same product be found (Directorate-General for Justice and Consumers (European Commission), 2018). In this paper, we propose the foundations of a new system for decentralizing product certificate based on blockchain technology which close the gaps of the unilateral published digital certificate. The decentralizing of product certification by using blockchain technology empowers consumers by easily verifying a product certificate, at the same time enable certification bodies to issue and sharing a certificate in tamperproof ways. This could be one of the important steps in the direction to reduce counterfeiting. The goal of this paper is to create a new category of applications in which the end-user identifies counterfeit product certificates and contribute to the fight against product piracy.
Article
Due to object detection's close relationship with video analysis and image understanding, it has attracted much research attention in recent years. Traditional object detection methods are built on handcrafted features and shallow trainable architectures. Their performance easily stagnates by constructing complex ensembles that combine multiple low-level image features with high-level context from object detectors and scene classifiers. With the rapid development in deep learning, more powerful tools, which are able to learn semantic, high-level, deeper features, are introduced to address the problems existing in traditional architectures. These models behave differently in network architecture, training strategy, and optimization function. In this paper, we provide a review of deep learning-based object detection frameworks. Our review begins with a brief introduction on the history of deep learning and its representative tool, namely, the convolutional neural network. Then, we focus on typical generic object detection architectures along with some modifications and useful tricks to improve detection performance further. As distinct specific detection tasks exhibit different characteristics, we also briefly survey several specific tasks, including salient object detection, face detection, and pedestrian detection. Experimental analyses are also provided to compare various methods and draw some meaningful conclusions. Finally, several promising directions and tasks are provided to serve as guidelines for future work in both object detection and relevant neural network-based learning systems.
Conference Paper
State-of-the-art object detection networks depend on region proposal algorithms to hypothesize object locations. Advances like SPPnet [7] and Fast R-CNN [5] have reduced the running time of these detection networks, exposing region pro-posal computation as a bottleneck. In this work, we introduce a Region Proposal Network (RPN) that shares full-image convolutional features with the detection network, thus enabling nearly cost-free region proposals. An RPN is a fully-convolutional network that simultaneously predicts object bounds and objectness scores at each position. RPNs are trained end-to-end to generate high-quality region proposals, which are used by Fast R-CNN for detection. With a simple alternating optimization, RPN and Fast R-CNN can be trained to share convolu-tional features. For the very deep VGG-16 model [18], our detection system has a frame rate of 5fps (including all steps) on a GPU, while achieving state-of-the-art object detection accuracy on PASCAL VOC 2007 (73.2% mAP) and 2012 (70.4% mAP) using 300 proposals per image. The code will be released.
Article
With the increase of globalization in trading and online shopping, phony products are on the rise in the market. This article introduces a range of technologies that have been implemented in the supply chain to deter counterfeiters. Technology for both product authentication and tracing and tracking products in the supply chain will be discussed, along with the advantages and disadvantages of each solution. Additionally, success stories on combating counterfeits are introduced. Finally, issues related to rising anti-counterfeiting costs, collaborative efforts in fighting fakes, and a multi-pronged plan are explored.
  • W Liu
Liu, W. et al., 2016. SSD: Single Shot MultiBox Detector. s.l., arXiv preprint, arXiv:1512.02325.
Yolov3: An incremental improvement
  • J Redmon
  • A Farhadi
Redmon, J. & Farhadi, A., 2019. Yolov3: An incremental improvement. s.l., arXiv preprint arXiv:1804.02767.
Global Brand Counterfeiting Report, s.l.: RESEARCH AND MARKERTS
  • Research
  • Markets
RESEARCH AND MARKETS, 2018. Global Brand Counterfeiting Report, s.l.: RESEARCH AND MARKERTS.
Anti-Counterfeit Packaging Market -Food And Pharmaceuticals by Authentication Packaging Technology
  • D Shields
  • R Deshmukh
Shields, D. & Deshmukh, R., 2020. Anti-Counterfeit Packaging Market -Food And Pharmaceuticals by Authentication Packaging Technology., 97220 United States: Global Opportunity Analysis and Industry Forecast.
Number of smartphone users worldwide from 2016 to 2021, s.l.: Statista
  • Statista
Statista, 2019. Number of smartphone users worldwide from 2016 to 2021, s.l.: Statista.
EEE transactions on neural networks and learning systems
  • Z Zhao
  • P Zheng
  • S Xu
  • X Wu
Zhao, Z., Zheng, P., Xu, S. & Wu, X., 2019. Object detection with deep learning: A review.. EEE transactions on neural networks and learning systems, 30(11), pp. 3212-3232. 18th International Conference e-Society 2020