Telepix, a total space AI solution company, announced on the 9th that six of its research results were accepted at the 'International Workshop on AI Foundation Model for EO in the Field of Earth Observation' co-hosted by the National Aeronautics and Space Administration (NASA) and the European Space Agency (ESA). Telepix was the only Korean company to have its research results selected at this workshop.
This workshop is an international event designed to discuss the development and utilization of technologies utilizing AI foundation models in the field of Earth observation satellite imagery, and is the first official event co-hosted by NASA and ESA. The event was held from the 5th to the 7th of this month at the European Space Agency (ESA) Earth Observation Centre (ESRIN) in Frascati, Italy.
The foundation model is a core technology in the AI field after deep learning, and is a general-purpose model that can learn from large-scale unstructured data and process various tasks. In particular, in the field of Earth observation where labeled learning data is insufficient, the foundation model is attracting attention because it can provide high analysis efficiency and accuracy.
Telepix presented a total of six research results at this event. One of these was presented orally and five were presented in poster format. They were evaluated and selected based on their practical applicability and innovation of satellite image analysis technology using AI.
In the oral presentation session, Jeong Young-sang, AI Solution Team Leader at Telepix, introduced ‘SatChat’, a satellite information-specialized AI chatbot based on a Large Language Model (LLM) developed by Telepix. SatChat is a user-friendly AI system designed to make it easy to search for satellite images and related information, and supports even non-experts to effectively utilize satellite data through multi-agent-based interaction functions. Since its official launch in December of last year, the system has also proven its commercial technology by signing a supply contract with a domestic public institution.
The poster presentation session introduced research results that are designed for practical service use, such as satellite image analysis techniques that increase the accuracy of raw material classification with limited data and a foundation model application method that can adapt to new satellite data. This is in line with Telepix’s direction to develop AI technology into a practical application stage.
“The fact that many of Telepix’s research results were adopted at international academic events jointly hosted by NASA and ESA is an example of the international competitiveness of Korea’s space AI technology,” said Kwon Darong-sae, head of Telepix’s data science division. “The results of this research will be applied to our satellite image analysis solution ‘Saetchat’ and raw material detection solution ‘Metalscope’, and will contribute to improving customers’ information analysis efficiency.”
- See more related articles
You must be logged in to post a comment.