LG AI Research Presents Research Achievements at CVPR 2022 and Hosts 'LG AI DAY'
In the first year of its establishment, LG AI Research had a total of 18 papers accepted by the global top-tier AI academic conferences in 2021. In its second year this year, LG AI Research presented 20 papers at prominent global academic conferences in just the first half of the year, thereby proving its research achievements in a short time.
Ranked the Top in the Field of Vision AI with the Independent Research Paper
LG AI Research showed its outstanding research outcomes by publishing a total of seven papers including an independent research paper at the “CVPR (Computer Vision and Pattern Recognition) 2022” that will be held at the Morial Convention Center in New Orleans, USA scheduled for six days starting from the 19th. In particular, the independent research paper (L-Verse: Bidirectional Generation Between Image and Text) describing the outcomes of “L-Verse,” which is a core technology of super-giant multi-modal AI that deals with both the language and visual information, was selected for presentation in the Oral Session. Only the top accepted papers are given an opportunity to present at the Oral Session.
CVPR is the world’s most authoritative academic conference in the field of computer vision, which is also called the “eye of AI.” It is a good opportunity to assess the achievements and level of research through this event. The fact that a company presents its independently researched paper at the Oral Session of the AI conference, which mainly presents studies conducted through industry-academia cooperation, represents that it has proven its technological competitiveness.
LG AI Research has also succeeded in commercializing super-giant multi-modal AI that allows bi-directional communication between language and visual information. The result was the 200 clothing collection unveiled by the AI artist Tilda in collaboration with designer Younhee Park during the New York Fashion Week earlier this year. Tilda interacted with designer Park to create over 3,000 images and patterns with the theme of “Flowers on Venus” for the collection.
Furthermore, LG AI Research is providing various types of services using multi-modal AI to its partners of “Expert AI Alliance,” which is a super-giant AI private sector alliance that was launched this year.
LG AI Research presented a total of seven papers comprising one workshop paper and six regular papers including L-Verse, which is the independent research paper, at this CVPR.
Advancing Toward the Center of Global AI Research
As can be seen from this CVPR, LG AI Research is making research achievements at a fast pace in just two years since its establishment. Earlier this year, it presented four studies at the internationally prominent AI academic conference ICLR (International Conference on Learning Representations), five studies at AAAI (Association for the Advancement of Artificial Intelligence), four at ACL (Association for Computational Linguistics), and seven studies at CVPR, thus presenting a total of 20 papers in global top-tier conferences in just the first half of the year. This is an astonishing feat as it already surpassed 18 papers presented in the entire year of 2021.
LG AI Research does not intend to just prove its current technological capacities but is looking to become a “global AI research hub” that will expand the research ecosystem globally. The company is expected to accelerate its R&D on AI platform technologies in the future. In fact, LG AI Research has more than doubled its research personnel in just one year since its establishment to reinforce its independent research capacities. Furthermore, it recently established the “SNU-LG AI Research Center” for joint research and human resource fostering on super-giant multi-modal AI with the Seoul National University Graduate School of AI.
It also founded LG AI Research Center, Ann Arbor, Michigan (US) last March to conduct research on AI advanced technology and to secure AI personnel in North America. It is conducting joint research with the University of Michigan as well as joint research for fundamental research to resolve difficult AI issues with the University of Toronto in Canada.
Starting with the LG AI Research Center, Ann Arbor, LG AI Research plans to expand its academia-industry collaboration with prestigious universities and research institutions in AI across North America and strengthen its open innovation.
Searching for AI Talent for Better Research
LG AI Research judged that while high-quality research infrastructure is also important, securing AI talent is even more crucial to achieving good research. Thus, it has been emphasizing its intent to continuously recruit competent AI researchers and has been engaged in various activities.
Its will to secure outstanding personnel will continue at the CVPR. In addition to LG AI Research, five affiliates such as LG Electronics, LG Display, LG Energy Solution, LG Uplus, and LG CNS will come to the conference in New Orleans to conduct joint AI talent recruitment activities during the event.
The five affiliates participating in this activity will operate a joint booth at the conference expo area for three days from June 21. The booth will demonstrate cases of LG overcoming difficult issues utilizing AI technologies in various industrial sectors, and it plans to conduct hiring meetings, etc.
To be more specific, LG AI Research will introduce technologies and cases for solving difficult problems using vision testing such as self-supervised representation learning, continual learning, and active learning. It will demonstrate DDU (Deep Document Understanding) in which AI reads and understands complex documents without human intervention and explain AI artist Tilda who participated in the New York Fashion Week together with a human designer, thereby suggesting the future of AI and human collaboration. Furthermore, the Fundamental Research Lab will introduce its research sectors and projects that the FR Lab is currently focusing on such as language understanding, multi-modal learning, molecular modeling, reinforcement learning, and continual learning.
The on-site booth will display a photo wall using images generated by EXAONE with the keyword of “jazz” that represents the New Orleans area. Affiliates including LG AI Research plan to construct a participation-type space where visitors freely add colors to the lines drawn by EXAONE to decorate the booth together to utilize this event as a space for more positive communication. 
LG X CVPR 2022
And on the 21st, LG AI Research and five affiliates will jointly host the LG AI DAY, which will be an event for LG employees and researchers participating in the conference to interact with each other. This is scheduled to be joined by about 200 academic researchers conducting the latest AI research from Korea, USA, Germany, etc. It is expected that this will serve as an opportunity for participants to build long-term relations through the banquet and networking event. In addition, the host including LG AI Research plans to utilize this event for joint research and to hire new employees. Kyunghoon Bae, the Representative of LG AI Research, stated, “We plan to develop our outstanding multi-modal AI technologies recognized by academia even further through research networks with the world’s top universities,” and added, “In addition to securing technologies, we will also focus on finding application cases beneficial to humans in order to realize the vision of LG AI Research to ‘Advancing AI for a better life.’”
Through this CVPR 2022, LG AI Research shared its astonishing research achievements and proved that its worldwide global research network is being operated successfully. It will also communicate actively with academia through an assortment of programs including experience-type booths and LG AI DAY, and continue to push toward its challenges to become a research institute that leads global AI technologies in the future.
▶[CVPR 2022] Large Vision-Language Model: What’s Next? (Link)
▶[CVPR 2022] L-Verse: Bidirectional Generation Between Image and Text - YouTube (Link)