TMCnet News
Superb AI Wins Global Vision AI Challenge at CVPR 2026From Fourth Place to First in a Single Year
SEOUL, South Korea, July 14, 2026 /PRNewswire/ -- Superb AI, an all-in-one Vision AI platform company led by CEO Hyun Kim (https://superb-ai.com/en), announced on June 18 that it took first place overall in the Foundational Few-Shot Object Detection Challenge at CVPR (Computer Vision and Pattern Recognition) 2026, the world's largest conference dedicated to computer vision. The challenge was held on June 4 as part of the Open-World Vision Workshop in Denver, with Superb AI's winning approach detailed in this technical report. Superb AI climbed from fourth place last year to the top spot this year, outperforming leading global teams backed by large-scale computing infrastructure through its efficient, industry-specialized methodology.
Held as part of CVPR's Open-World Vision Workshop, the challenge evaluates few-shot object detection—the ability of an AI model to identify new objects using just 10 example images per object class. It is considered a demanding benchmark for determining whether AI can be deployed directly in industrial settings without the extensive data collection and labeling typically required. This year's challenge used Roboflow20-VL FSOD, a dataset covering 20 specialized domains rarely represented in general internet data, including X-ray, thermal, and aerial imagery. Using its proprietary, industry-specialized Vision Foundation Model ZERO, Superb AI achieved an mAP (mean Average Precision) score of 53.9 across the 20 domains. The result surpassed the 51.6 recorded by the second-place joint team from China's Fudan University and Lenovo, marking Superb AI's rise from fourth place in the same challenge last year to first place in just one year. It also significantly exceeded the organizers' baseline model score of 33.3. Superb AI ranked first in five of the seven categories spanning 20 diverse domains: Aerial, Docs, F&F, Industry, Medical, Sports, and Other. The company secured a commanding first-place finish in the Industry category with a score of 64.4. In the Medica category, it scored 51.4—more than nine points ahead of the runner-up. By sustaining top-tier performance across a broad range of distinct and heterogeneous settings, rather than excelling in only one domain, ZERO demonstrated its practical industrial value: a single model can be rapidly adapted to diverse environments using limited data. ZERO, the model used by Superb AI in the challenge, is the company's proprietary, industry-specialized Vision Foundation Model. Pretrained on large-scale image data, ZERO provides strong zero-shot generalization as its foundation. For the challenge, Superb AI efficiently adapted the model to each new domain using only the 10 labeled examples provided per object class. By reducing the time and cost required to adopt AI, the model enables rapid deployment across a wide range of industries, including manufacturing, logistics, and healthcare. Last year, Superb AI focused on validating ZERO's core capabilities. This year, the company concentrated on building a framework that enables the model to adapt quickly and efficiently to individual industry domains. Competing against Chinese research institutions backed by government-supported, large-scale data and infrastructure, Superb AI took the top spot through its industry-specialized strategy and efficient methodology. "This win is about more than receiving an award. It shows that an industry-specialized strategy and efficient methodology can reach the highest global level without relying on costly, large-scale infrastructure," said Hyun Kim, CEO of Superb AI. "We will continue to focus on building AI that works in real-world industrial environments—not just on achieving strong benchmark results." "What matters most about this achievement is that it brings research outcomes and industrial application together," said Moonsu Cha, CTO of Superb AI, who led the research team. "ZERO is not designed simply to score well. It is designed to be a practical model that can be deployed quickly and efficiently in customer environments." (End) - Superb AI: An All-in-One Solution Provider Supporting Every Stage of Vision AI Adoption Founded in 2018, Superb AI is an MLOps (machine learning operations) startup supporting the full AI development lifecycle. The company provides a broad portfolio of products and services that help enterprises build and successfully deploy AI systems. Its solutions enable customers to establish a complete, iterative MLOps pipeline spanning dataset creation, selection, processing, management, and analysis through model training and deployment. Since its founding, Superb AI has focused on helping enterprises adopt Vision AI using images, video, and 3D LiDAR data. Its range of intuitive solutions enables users to develop and manage AI even without coding or machine learning expertise. The system is streamlined enough to allow users to build deep learning models with just a few clicks. In 2025, Superb AI launched Superb AI Video Analytics, a video-based AI solution that detects anomalies—including fires, falls, and intrusions—in real time across multiple CCTV feeds while also providing intelligent text-based search. The solution enables advanced monitoring automation across smart cities, industrial sites, logistics facilities, and public institutions. Powered by VLM-based multimodal AI, it delivers greater accuracy and usability than conventional video monitoring systems. The same year, the company introduced ZERO, its Vision Foundation Model recognized as Korea's first practical AI model designed for immediate deployment in industrial environments. ZERO enables users to perform desired tasks using only text prompts or example images, without complex coding or machine learning expertise. Its zero-shot capabilities also allow it to adapt flexibly to new environments without additional training. Superb AI provides AI solutions to more than 100 enterprise customers worldwide, including leading technology companies such as Toyota, KT, Samsung, LG Electronics, Qualcomm, Hyundai Motor Company, and SK Telecom. The company operates entities in Korea, the United States, and Japan as it continues to expand its global presence. To date, Superb AI has raised approximately US$ 41 million in cumulative funding from investors in Korea and the United States. The company has obtained SOC 2 Type II certification, demonstrating its adherence to rigorous global security and operational standards, and holds 17 patents in Korea and the United States. Superb AI filed its first Korean patent just one month after its founding and has since registered patents related to the auto-labeling and custom auto-labeling capabilities of Superb Platform. In recognition of these achievements, Superb AI was selected in December 2022 as the top-performing company in the Seoul Metropolitan Government's R&D program based on its patent registration record. Superb AI offers Superb Platform, a high-performance AI development platform that supports the entire AI lifecycle within a single system. The company also provides Superb Service, through which its experts deliver end-to-end support spanning consulting, data design, AI algorithm development, and operations. This service draws on Superb AI's extensive experience across customer projects in diverse industries and its expertise in building production-grade, high-performance AI systems. In addition, Superb AI offers vertical AI solutions tailored to industrial environments, including Superb AI Video Analytics, Superb AI Defect Detection, and Superb AI Edge. These solutions can be combined and delivered according to each customer's needs.
SOURCE Superb AI, Inc. 
|
View original content to download multimedia: