On 25^th, SNU College of Engineering (Dean Cha Kook-Heon) announced that Professor Zhang Byoung-Tak’s team “SNU-BI” (consisting of PhD Candidate Kim Jin-Hwa and MS Student Jun Jaehyun) of the Department of Computer Science and Engineering has won second place at the 2018 VQA (Vision Question Answering) Challenge, a workshop of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR), that was held at Salt Lake City on 18^th.
 
Out of the 40 teams participating, Professor Zhang’s team demonstrated highest excellence as a single model, but slightly lacked as an ensemble model to finish second by a narrow margin.
 
First round being held in 2016, the VQA Challenge is an annual event where enhanced technologies are staged and evaluated. From 2017, the VQA v2.0 train collects different answers replied on a single question provided with different images to strictly assess the effectiveness of artificial intelligence (AI).
 
For instance, when a similar image is provided on the question of “Who is wearing glasses,” the data set restricts the distribution of possible answers to “Man” and “Woman.” The VQA v2.0 train contains more than 200K images and 1.1M questions, which is at least 1.8-times improvement from the VQA v1.0 train.
 
The final results of this challenge can be found on VQA Challenge homepage (http://visualqa.org/challenge.html).
 
This technology developed by Zhang’s team is a result of the AI National Strategy Assignment Video Tutoring Test Project (VTT) and StarLab Project of the Ministry of Science and ICT.