Naver Presents 13 Papers at International Conference on Computer Vision... Attention on Service Application Potential View original image


[Asia Economy Reporter Kang Nahum] Naver announced on the 18th that it presented a total of 13 papers at the International Conference on Computer Vision (ICCV).


Now in its 18th year, ICCV is regarded as the premier conference in the field of computer vision. This year, it was held online from the 11th to the 17th. Naver has participated in ICCV since 2017, and as it has strengthened its investment in artificial intelligence (AI), the number of accepted papers has increased every year.


At this conference, Naver, together with Naver Labs Europe and LINE, presented a total of 13 papers in the regular sessions. Among them, one paper earned the opportunity to present in the oral session, which is awarded to approximately the top 3% of research.


Naver presented many papers at this conference that are not only meaningful long-term preliminary studies but also applicable to actual services. Notably, the paper titled "Multiple Heads are Better than One: Few-shot Font Generation with Multiple Localized Experts" proposes a technology that can efficiently generate fonts with much less handwritten data than before. Naver Clova has actively incorporated related AI technologies into actual services, such as introducing 109 types of ‘Nanum Handwriting’ fonts created from user handwriting using AI technology in 2019.


Additionally, the paper "Rethinking Spatial Dimensions of Vision Transformers" presents a method to model Vision Transformers, which have recently gained attention for outperforming conventional Convolutional Neural Network (CNN) models in image recognition, more efficiently. It is expected to be utilized to enhance various Naver services where image recognition is applied by demonstrating high accuracy even with limited data.


The paper "Keep CALM and Improve Visual Feature Attribution," which deals with ‘trustworthy AI,’ studied an interpretation technique that can improve the bias of CAM (Class Activation Mapping), widely used for interpreting image recognition results. This research was conducted jointly with the University of T?bingen in Germany. In July, Naver declared cooperation with the University of T?bingen to develop safe and trustworthy AI research.



Ha Jungwoo, Head of Naver Clova AI LAB, said, "AI is evolving from research to a technology that creates new value in people's daily lives," adding, "Naver has been making extensive investments in AI technology, from theory to research applicable to actual services, and continues to achieve results."


This content was produced with the assistance of AI translation services.

© The Asia Business Daily(www.asiae.co.kr). All rights reserved.

Today’s Briefing