Korean startups capture niche AI market in image, video analysis

Graphics by Kim Eui-gyun, DALL-E

South Korean startups specializing in image and video artificial intelligence (AI) are gaining notable success worldwide, focusing on niche areas of AI that have traditionally been dominated by American giants.

Luxury conglomerate LVMH is using AI to detect counterfeit versions of its high-end brands. The AI system continuously scans e-commerce sites and social media, automatically identifying fake luxury goods, collecting evidence, and reporting violations. This technology was developed by MarqVision, a startup founded by Korean entrepreneur Lee In-sup in the United States in 2019. A former McKinsey consultant, Lee has built MarqVision into a company serving 300 global brands and content businesses worldwide. Its AI detects not only counterfeit goods but also unauthorized product sales and pirated content, such as illegal webtoons and movies. Last year alone, the company identified 50.41 million cases of counterfeits and illegal sales, generating $17 million (about 25 billion won) in annual global revenue.

While large language models (LLMs) are dominated by well-funded U.S. tech giants like OpenAI, Google, Microsoft, and Meta, Korean startups with deep expertise in visual AI hold a competitive edge in specialized industries. Experts predict that AI-driven image and video technologies will become essential in fields like autonomous driving.

Twelve Labs, the first Korean startup to receive major investment from AI chipmaker Nvidia, has gained attention for its AI-powered video search technology. The AI can locate a specific scene, such as “a man holding a pen in an office,” within hours-long footage in just one second. North America’s largest sports entertainment firm, MLSE, and the National Football League (NFL) already use Twelve Labs’ AI to analyze sports footage, retrieve key moments, and generate new content.

Korean AI startups are also making strides in medical imaging, where non-language data like MRI and CT scans are critical. Medical AI startup AIRS Medical has developed software that enhances MRI resolution, speeds up scan times, and restores image quality. Its technology is used in 460 medical institutions across 26 countries, including the U.S., U.K., Germany, and Japan, with more than half of its revenue coming from overseas. Another startup, Inocras, has developed AI that analyzes a person’s entire genome for cancer diagnosis and personalized treatment. The company is working with the Broad Institute, a genomics research center affiliated with MIT and Harvard, to reanalyze data from the world’s largest public cancer patient database, The Cancer Genome Atlas (TCGA).

Graphics by Kim Eui-gyun

Korean startups' success in non-language AI can be attributed to the relatively lower development difficulty compared to language models. Language models rely heavily on extensive text data, mostly in English, which poses challenges for countries with limited access to such data. Lee Kang-joon, CEO of Dunamu & Partners, which invested in AIRS Medical and Inocras, explained, “Korea benefits from having centralized medical data from Seoul’s five major hospitals and the Health Insurance Review & Assessment Service, whereas in the U.S., developers must collect data from multiple hospitals across the country, making development more complex.” Specializing in industry-specific AI models has also proven beneficial. A representative from MarqVision noted, “In Asian countries with a well-developed e-commerce sector, gathering counterfeit image data for training is more advantageous. Our AI technology, developed long before the emergence of ChatGPT, is now being recognized internationally.”

[Ahn Sang-hyun]

- Copyrights ⓒ 조선일보 & chosun.com, 무단 전재 및 재배포 금지 -