□ DGIST (President Kunwoo Lee) has achieved global recognition through industry-academia collaboration, being selected as a final presenter for the Industry Day Talks, a core session of CIKM 2025, one of the world’s most prestigious conferences in the field of artificial intelligence (AI).
□ CIKM is a major international academic conference in data mining, information retrieval, and AI, serving as a global venue for exchange among world-class companies such as Google, Meta, OpenAI, and Amazon. This year, only 13 organizations worldwide were granted presentation opportunities at the Industry Day Talks. The DGIST–TORIS team was the only team to be selected from Korea and from all of Asia, clearly demonstrating their technological excellence.
□ The presentation showcases an anti-drone AI technology jointly developed by Professor Ji-woong Choi’s team from DGIST’s Department of Electrical Engineering and Computer Science and TORIS, a start-up founded by Dr. Daegeon Oh, Senior Researcher at DGIST’s Intelligent Robotics Research Division. The research is highly significant, as it is the world’s first technology to extend the identification range of small drones from the previous limit of 2–3 km to 8 km.
□ Previous technology lacked high-performance radar and infrared (IR) camera systems capable of detecting and tracking small targets at such long distances, making it impossible to collect the training data required for AI-based long-range identification, which posed a fundamental limitation to technological development. Moreover, at such distances, small drones appear only as dots, making shape-based recognition infeasible, and existing AI models often failed when encountering new drone types.
□ To overcome these challenges, the DGIST–TORIS research team introduced a novel AI approach focused on movement. Unlike birds, drones exhibit stable and repetitive flight patterns. The team designed an AI system capable of learning and distinguishing such differences in movement, including wing-flapping patterns. They trained a 3D Convolutional Neural Network (3D-CNN) using long-range video data collected through TORIS’s world-class integrated radar–infrared (IR) camera system.
□ As a result, the research team successfully identified small drones at a record-breaking distance of 8 km, a result never reported in academic literature, and achieved stable performance even in low-resolution environments. The system accurately recognized newly encountered drone models and significantly reduced false alarms caused by non-threatening objects such as birds.
□ This accomplishment is considered a world-class technological breakthrough that will dramatically enhance long-range anti-drone defense capabilities in military and security sectors. It is also garnering strong attention as a potential new export item for Korea’s K-Defense industry.
□ Daegeon Oh, CEO of TORIS, stated, “This achievement demonstrates that the combination of DGIST’s AI technology and TORIS’s world-class hardware can solve complex challenges in defense and security. We are proud to present our technology on a global stage alongside companies like Google and OpenAI.” Professor Ji-woong Choi of DGIST added, “This research is a huge success story of collaboration among academia, research, and industry, and it will serve as great motivation for our students. We will continue pursuing research that contributes to the advancement of the nation and regional industries.”
□ Meanwhile, this research outcome stems from the synergy between DGIST’s AI capabilities and TORIS’s cutting-edge radar and imaging technologies. First-author Woochul Jin, a student participating in DGIST’s industry–academia cooperation program, worked directly at TORIS on radar and IR sensor development, laying the foundation for the study’s success. Co-corresponding authors include Professor Ji-woong Choi and Senior Researcher Sang-Chul Lee of DGIST (concurrent Nano-Convergence Technology Division AI major), with TORIS CEO Daegeon Oh also contributing to the research.
