-  1^st place in 3 events for 2 consecutive years


▲ SNU Professor Kim Kun-Hee’s team in ICCV 2017

 
On October 31^st, SNU College of Engineering (Dean Cha Kook-Heon) announced that SNU Department of Computer Science and Engineering Professor Kim Kun-Hee’s team has won LSMDC (Large Scale Movie Description Challenge), which is a workshop of ICCV (International Conference on Computer Vision) held in Venice, Italy.
LSMDC is a contest that evaluates the ability of artificial intelligence to describe movies after watching them for less than 10 seconds, or to create appropriate answers when asked questions in natural language. Professor Kim’s team (Yoo Young-Jae, Kim Jong-Seok) has participated in all 4 events and won 3 (Movie Annotation and Retrival, Movie Multiple-Choice Test, Movie Fill-in-the Blank) and 4^th place in 1 event (Movie Description).
Prof. Kim’s team participated and won same 3 events last year, and is maintaining world-class technology in machine-learning algorithm in movie question-answering.
They also participated in MovieQA, which is a QA contest on movie story comprehension and won 2^nd place. This contest evaluates the ability of artificial intelligence to understand a new movie and answering questions about the movie. It’s a challenging task requiring comprehensive understanding of the movie, and Professor Kim’s team (Na Se-Il, Lee Sang-Ho) has won the 2^nd place with a narrow margin of less than 1% with the 1^st place Tianjin University team of China. The Deep-learning algorithm used in this contest was published as a paper in ICCV 2017.
Both contests evaluate the ability of artificial intelligence to understand and describe videos in languages that human can understand, or to find answers to questions asked by humans, requiring high-level artificial intelligence on visual perception and natural language processing. This technology can be used in movie-explanation service for the blind, or in natural-language searches of movies and discussions with artificial intelligence on movies.