Search across all content
projectReading accuracy to detect illegal duplication is expected to improve by 93% from 0.4% of the existing visual inspection in the case of introducing AI solution.
This information was sourced from "AI Convergence Reading System for Illegal Duplication Project." by Observatory of Public Sector Innovation (OECD), https://oecd-opsi.org/innovations/ai-convergence-reading-system-for-illegal-duplication-project/. Licensed under Creative Commons Attribution-ShareAlike 3.0 IGO (CC BY-SA 3.0 IGO).
As South Korea, the 9th largest trade partner has increased K-brand power due to the spread of the Korean Wave to the world recently, damage cases in IPRs have accordingly increased from the inflow of illegal duplication imitating domestic products. The number of detecting goods infringed IPRs in the process of customs clearance in 2019 was 30,856 which accounted for 0.4 billion dollar annually. South Korea falls victim of the top 8th country by illegal duplication and see 15 billion dollar dropping per annum in company sales, 31,753 people in job losses, and 0.3 billion dollar in tax income leakage.
In the case of distributing illegal duplication, companies have the sunk cost of IPRs/R&D and suffer losses on a large scale due to losing market opportunities. For consumers, they suffer direct health and safety damage including ingesting fake medicines and appliance fires. For example, Hyundai Mobis experienced damage of 18 million dollar annually due to domestic and overseas distribution of cheap Chinese fake ‘sensors, infotainment, steering devices, lamps, body parts. etc.’. In the case of failing to detect in customs clearance, the domestic distribution expands on a large scale.
Initial data is required for AI reading of major customs clearance items in 2020 but new business project faces challenges due to rejecting to provide company’s design rights of drawing (business secret) and lacking design information about imported items in customs clearance. Patent-holding companies hesitate to participate in the National IT Industry Promotion Agency (NIPA) AI demonstration-supported project for concerns about technical leakage in providing information of design rights (including goods names, etc.). Korea Customs Service’s cooperation is crucial to AI model demonstration in customs clearance environment needed for a high level of security /quarantine, but the ministry lacks credibility for the initial AI solution.
NIPA and Korea Customs Service empathize with damage impacted on industry and people in 2020 in the case of the distribution of illegal duplication and have been developing AI-based reading solution to detect illegal duplication utilized customs clearance data through collaboration.
Starting with a demand survey for AI convergence tasks by inter-ministerial collaboration in February 2020, we made detailed business project proposal jointly through a number of consultations with NIPA-the Ministry of Science and ICT-Korea Customs Service. For smooth securing customs clearance data (image of actual goods, patent of design rights, etc.) which is essential for AI solution, business promotion & presentation are held every year for more than 100 companies of IPRs in cooperation with Korea Customs Service. NIPA as dedicated institution is working with multiple institutions such as the Ministry of Science and ICT as hosting department, Korea Customs Service as demand department as well as Trade-Related Intellectual Property Promotion Association (TIPA) as IPRs-holding company groups.
We secure reading accuracy and improve the efficiency of customs clearance by changing reading method to detect illegal duplication based on AI & customs clearance data from the existing visual inspection. We will achieve 93% of reading accuracy by 2023 and develop the solution to apply Incheon & Pyeongtaek Customs where Chinese items are imported most.
We achieved 82.7% of reading accuracy which is above 80% of our target for initial AI solution models developed by 2021. We expect to prevent infringement of IPRs by 60% (0.24 billion dollar per year compared to 0.4 billion dollar per year) of the overall inflow amount through illegal duplication in the case of applying to Customs in 2024 by accelerating the solution by 2023. Reading accuracy to detect illegal duplication is expected to improve by 93% from 0.4% of the existing visual inspection in the case of introducing AI solution. We can reduce the time to notify the right holder after detecting initial suspected goods, less than 5 minutes in the case of introducing AI solution from more than 1,440 minutes through visual inspection.
This AI system achieves following two major innovations.
As of 2021, the AI system can read pirated copies with 82% reading accuracy. In addition, it is applied to the customs of Incheon and Pyeongtaek in Korea and is being tested to detect illegal copies entering from China. By 2022, the accuracy of reading illegal copies of artificial intelligence systems is expected to improve to 90%.
In addition, we are contributing greatly to the protection of intellectual property rights in Korea's strong industries by applying artificial intelligence piracy reading system to 89 major products that Korea has global competitiveness such as automobile parts, IT products, household goods, and rice products.
By 2023, we plan to increase the reading accuracy of the AI system to 93% or more, and from 2024, we plan to apply this AI system to all customs offices in Korea.
This AI project began in 2020 jointly organized by the Ministry of Science and ICT and the Korea Customs Service. The Korea Customs Service was having a hard time reading the explosive increase in illegal copies during customs clearance. The Ministry of Science and ICT found that if artificial intelligence and customs data can be fused to automatically read illegal copies, it can significantly reduce the burden on customs inspectors. NIPA also draws collaboration from companies with IPRs for AI.
For companies with IPRs in Korea, we expect to prevent infringement of IPRs by 60% (0.24 billion dollar per year compared to 0.4 billion dollar per year) of the overall inflow amount through illegal duplication in the case of applying this AI system to Customs in 2024.
For Customs officers, we can reduce the time to notify the right holder after detecting initial suspected goods, less than 5 minutes in the case of introducing AI solution from more than 1,440 minutes through visual inspection.
As of 2022, the AI system is being tested at two major customs offices in Korea. AI system is used to examine automotive parts, IT products, and beauty products as detailed items. The Korean government predicts that the accuracy of reading illegal copies of AI system will reach 90% by 2022. In addition, AI system reading accuracy is expected to improve to 93% by 2023 (currently 82%).
Korea imports about $600 billion worth of products from overseas annually. The Korean government predicts that about 2.2 percent of total imports will be illegal copies. If the AI system is applied to 34 customs in Korea from 2024, it is expected that up to 60% of the import amount of illegal copies will be detected when customs clearance is performed. The AI system not only protects companies with intellectual property rights in Korea from illegal copies, but also can increase the level of safety of the people by blocking imports of fake medicines and auto parts, for example.
There were three success factors before the AI system was successful in Korea.
This AI system has various scalability. AI robots can search for fake products in online shopping malls to find illegal copies. In addition, companies can use this AI system to identify similar product designs in advance at the stage of applying for patents for intellectual property rights in Korea, such as trademarks and design rights. This allows companies to check in advance whether the patent they apply for violates existing intellectual property rights.
People can also use this AI system in our daily lives. Consumers can always check whether the product they purchase and use at an online shopping mall is genuine or illegal. The AI system is expected to provide a wide range of usability to businesses and individuals.
In addition to the customs clearance field presented in this case, AI can be widely used in areas close to public safety, such as medical care, finding missing children, and detecting harmful chemicals at chemical disaster sites. The Ministry of Science and ICT and the National IT Industry Promotion Agency (NIPA) have been developing AI services in earnest since 2020. AI is applied to various areas, ranging from the public safety area to the detection of defective products at manufacturing sites that Korea has advantages, and productivity improvement.
Through this convergence of various industrial areas of AI, it was confirmed that productivity in various fields such as manufacturing, finance, and medical care was greatly improved. On the other hand, we have learned that if data containing people's personal information is acquired during this process, these data also need to be managed reliably in a safe way.
We are very happy to introduce the artificial intelligence illegal reproduction reading system that is applied to customs clearance sites in Korea. Korea's artificial intelligence system is a project being promoted by the government and companies with intellectual property rights. Although there were difficulties in securing data in the early stages of the project, we have now achieved results such as securing more than 82% of illegal copies. And in 2024, we plan to apply this system to 34 customs sites in Korea.
I hope there will be an opportunity to share Korea's experience in developing an artificial intelligence illegal reproduction reading system and applying customs demonstration to OECD member countries.
Budget: Not disclosed
Connect with 500,000+ public servants solving your hardest challenges.
Connect with 500,000+ public servants solving your hardest challenges.
Help public servants worldwide learn from your work, what worked, what flopped and what you'd do differently
Share your project
Log in or sign up to continue the conversation