LG presented two research papers at the prestigious International Conference on Learning Representations (ICLR) 2024 in Vienna. In detail, the research paper “DiffMatch: Diffusion Model for Dense Matching” introduces an AI technology adept at comparing images to pinpoint similarities and variations, facilitating precise identification of object locations and shapes. This innovation holds promise for enhancing spatial recognition speed and accuracy in advanced robotics applications, allowing robots to navigate environments with changing object positions and lighting conditions efficiently. In addition, the paper titled “H2O-SDF Two-Phase Learning for 3D Indoor Reconstruction using Object Surface Fields”, which focuses on a core technology for the metaverse, illustrates the utilization of AI to reproduce complex indoor spaces and details of object surfaces in a 3D virtual space. Using 2D images as source material, the AI can identify and render an interior structure – including walls, ceilings and pillars – in 3D. Following this, the technology leverages learned information to add fine details to furniture pieces, home appliances and other household objects, bringing a high degree of realism to virtual spaces. Suitable for diverse applications relating to the metaverse, LG’s innovative tech can also potentially be used to augment existing “digital twin” software, which has become an essential tool for smart factories. “By applying our own cutting-edge AI technologies to LG products and services, we will continue to make customers’ lives more convenient and enjoyable – whether it’s at home, at work or in the virtual realm”, said Kim Byoung-hoon, Chief Technology Officer at LG Electronics.
