Researchers have made a breakthrough in artificial intelligence vision technology, enabling robots to swiftly and accurately identify objects they’ve never encountered before.
On Monday, Gwangju Institute of Science and Technology (GIST) announced that Professor Lee Kyu Bin’s research team in the AI Convergence Department has developed a new AI technology that refines object recognition by incorporating error estimation.
This innovative system can detect and correct misidentified objects in real time, significantly enhancing a robot’s visual perception capabilities.
Current AI vision technologies face several limitations, including the inability to recognize objects not previously trained and reduced accuracy in complex environments without human intervention.
The research team developed the QuBER model to address these challenges, which employs rapid and precise error correction.
QuBER analyzes Quadruple Boundary Errors using RGB-D (Red-Green-Blue-Depth) images and initial prediction data, improving object recognition accuracy. This groundbreaking technology allows robots to identify previously unseen objects with speed and precision in real time.
The QuBER model has demonstrated world-class accuracy, showcasing impressive segmentation capabilities even in challenging scenarios where multiple objects are obscured.
Professor Lee Kyu Bin expressed enthusiasm about the research, saying, “This study confirms that robots can accurately and efficiently recognize unfamiliar objects. We anticipate this technology will be instrumental in developing robots capable of reliably performing various tasks in new environments.”
The research, led by Professor Lee Kyu Bin and conducted by PhD student Baek Seung Hyuk, was supported by the Ministry of Trade, Industry and Energy and the Ministry of Science and ICT.
The team is scheduled to present its findings at the International Conference on Robotics and Automation (ICRA) in May 2025, one of the most prestigious conferences in the robotics field.
