Of all the senses, vision provides the most information to humans. Humans also make sense of the world through language, which abstracts and symbolizes a lot of information. In this way, visual and linguistic information are completely different in form, but are linked together in an abstract space to provide human cognition. With this in mind, the Multimodal Lab unravels and models the information entangled between vision and language data, opening up new possibilities. Multimodal lab is developing AI model, which uses multimodal AI to generate images from text and vice versa. We are also challenging new translational medicine technologies by fusing medical imaging data with heterogenous clinical/test data.
Multimodal ModelWe are conducting multimodal research and developing application technologies for integrated understanding of data across different types of data, especially 2D and 3D visual data and language data. We study image generation and image captioning technologies through generative models, prompt engineering and model optimization technologies to improve performance in real-world applications, and research vision foundation models as the basis for various vision tasks.
Medical Data AnalysisWe are aiming to build a Medical Foundation Model that encompasses various medical imaging (Radiology, Tomography, Microscopy) and clinical
and various test data. We are focusing on the development of multimodal models that can perform various medical tasks based on refinement and
processing technologies for vast and heterogeneous medical data, presenting a new horizon for translational medicine.
Meet our leader!
Soonyoung Lee, Head of Multimodal Lab
“Humans acquire diverse and complex external environmental information primarily through vision, and understand the world by interpreting visual information through language, an abstract and symbolic means. Visual information and verbal information are completely different in form and characteristics, but they are interconnected at an abstract level and form human cognitive abilities. Multimodal Lab takes note of these observations and interprets and models the complex interactions that exist between visual and verbal data, opening up new opportunities and possibilities.
We want to explore the innovative possibilities of generative artificial intelligence in various fields, implement them, and turn them into reality. If you’re full of new ideas and passion, join us in making a real change int he world.”