Nota (CEO Chae Myung-soo), an AI lightweight and optimization technology company, announced on the 19th that it has completed KBS's "Disaster CCTV AI Dataset Construction and Video Analysis Enhancement" project. Through this project, Nota implemented an on-device AI-based disaster news special workflow utilizing a Vision Language Model (VLM).
The business goal was to overcome the limitations of the existing disaster broadcasting process, which required manual review of numerous CCTV footage, and to ensure the speed and accuracy of disaster broadcasting by having AI select footage optimized for broadcasting in real time.
For this project, Nota utilized its "NVA (Nota Vision Agent)" as the core engine. NVA efficiently drives high-performance VLMs, analyzes CCTV footage from disaster areas, and automatically selects scenes suitable for broadcasting. Furthermore, by incorporating the opinions of reporters conveying disaster information, Nota established "newsworthiness" as a selection criterion, enhancing practical usability.
Internal testing results showed a high level of consistency between the images selected by the NVA and those selected by journalists from the wildfire dataset. Even with massive data processing, the system operated reliably, ranking images optimized for disaster situations and providing a basis for decision-making within tens of seconds. This created an environment that could dramatically shorten the time it takes to deliver disaster information to the public.
KBS has established a user feedback system that allows reporters to evaluate the appropriateness of their reporting, and this feedback is reflected in improving NVA performance. Nota is gradually improving the system based on field feedback and feedback data.
Nota CEO Chae Myung-soo said, “This project is a meaningful example of AI optimization technology and on-device AI capabilities expanding into the broadcast media sector,” and added, “We will continue to develop AI technology that can be helpful in disaster response sites through collaboration with the public and media sectors.”
A KBS official explained, “We have implemented a system that can be used immediately by combining disaster broadcasting experience and AI optimization technology,” and “It is significant in that we have established a new infrastructure that can accurately select videos suitable for disaster special reports using VLM and deliver information to the public.”
- See more related articles
You must be logged in to post a comment.