The paper 'ResNet-Based Birdsong Detection with a High-Pass Filter', by graduate students Lee Kang-woo (first author), Kwon Tae-hyun, Ka Sang-min and Kim Yoon-young of the Department of Computer Science & Engineering (supervisor: Professor Jung Sung-won), received the Best Paper Award at the 2022 Summer Workshop on Computer Communications (SWCC 2022) held on 26 August.

▲ From left: Lee Kang-woo (master's, 3rd semester), Kwon Tae-hyun (master's, 3rd semester), Ka Sang-min (master's, 2nd semester), Kim Yoon-young (master's, 1st semester) and Professor Jung Sung-won (supervisor)

▲ Best Paper Award at the 2022 Summer Workshop on Computer Communications (SWCC 2022)

BirdCLEF 2022 (Kaggle) is a competition to identify 21 bird species from their calls. Given 5,500 one-minute test recordings, the goal is to classify each five-second segment as 'call (true)' or 'nocall (false)' for each of the 21 species. The paper identifies three problems with the data provided in BirdCLEF 2022, including weak labelling, and as a solution carries out four stages of data preprocessing: data augmentation, min-frequency extraction, high-pass filtering, and hand classification to remove 'nocall' segments. It then builds a ResNet-34 model, which performs well on image classification, trains it on the preprocessed dataset, and achieves 97% on the F1 score.

The 2022 Summer Workshop on Computer Communications (SWCC 2022) is a forum for active exchange among academia, industry and research institutes, at which key issues in information and communications are selected and discussed in depth by relevant specialists.