MayAI Honored with ICLR 2025 Top 5% Spotlight Paper

Mayi has globally proven its innovative AI technology by being selected for the 'Spotlight' grade at ICLR 2025, the world's most prestigious artificial intelligence academic conference. This is a significant achievement that further strengthens its position as a global AI company. AI video processing startup Mayi earned the honor of being selected for the 'Spotlight' grade, awarded to the top 5% at ICLR 2025 (International Conference on Learning Representations), one of the top three global AI conferences, with its paper titled 'Research on Camera Bias in Person Re-identification AI Models'. ICLR is recognized as a top-tier academic conference in the AI field, alongside NeurIPS and ICML (International Conference on Machine Learning). This paper, co-authored by researchers Song Myung-seo, Park Jin-woo, and Lee Jong-seok, analyzed the limitation that existing AI models struggle to re-identify the same person due to differences in CCTV environments, and proposed an innovative solution to overcome this. The research team, through improvements in 'normalization techniques to reduce camera bias' and 'unsupervised person re-identification technology learning methods', implemented the ability to maintain consistent accuracy across various conditions and correctly recognize people even in new, unlearned spaces. This is a core technology that can maximize operational efficiency for various stores by achieving high accuracy with minimal human resources. In particular, Mayi's self-developed re-identification (Re-ID) technology demonstrated its excellence by achieving 92% accuracy in internal tests, surpassing the latest academic model (ISR, 66%). This core technology is already applied to Mayi's solution 'mAsh', contributing to precisely analyzing visitor data in offline spaces and maximizing store operational efficiency. Song Myung-seo, lead of the research team and first author of the paper, stated, "I am very pleased that our efforts to improve mAsh's accuracy have been recognized globally," and added, "Through this achievement, we expect to be able to provide even more reliable and precise analytical data to our customers." Kim Chan-gyu, CPO of Mayi, stated, "The person re-identification model recognized this time is Mayi's core capability, allowing for precise analysis of store visitor data while prioritizing personal information protection." He expressed his ambition, "Moving forward, we will continue to innovate based on world-class technology to provide customers with the most accurate and reliable data." Previously, in September last year, Mayi obtained a patent for its AI-based video analysis technology from the U.S. Patent and Trademark Office, thereby strengthening the legal protection of its offline store customer behavior analysis technology and solidifying its global market position.