Jaehun Park

Jaewon Son, Jaehun Park, Kwangsu Kim

Video summarization aims to generate a concise representation of a video, capturing its essential content and key moments while reducing its overall length. Although several methods employ attention mechanisms to handle long-term dependencies, they often fail to capture the visual significance inherent in frames. To address this limitation, we propose a CNN-based SpatioTemporal Attention (CSTA) method that stacks each feature of frames from a single video to form image-like frame representations and applies 2D CNN to these frame features. Our methodology relies on CNN to comprehend the inter and intra-frame relations and to find crucial attributes in videos by exploiting its ability to learn absolute positions within images. In contrast to previous work compromising efficiency by designing additional modules to focus on spatial importance, CSTA requires minimal computational overhead as it uses CNN as a sliding window. Extensive experiments on two benchmark datasets (SumMe and TVSum) demonstrate that our proposed approach achieves state-of-the-art performance with fewer MACs compared to previous methods. Codes are available at https://github.com/thswodnjs3/CSTA.

Thanh-Cong Dinh, Hyerim Bae, Jaehun Park et al.

In this paper, we study idea mining from crowdsourcing applications which encourage a group of people, who are usually undefined and very large sized, to generate ideas for new product development (NPD). In order to isolate the relatively small number of potential ones among ideas from crowd, decision makers not only have to identify the key textual information representing the ideas, but they also need to consider online opinions of people who gave comments and votes on the ideas. Due to the extremely large size of text data generated by people on the Internet, identifying textual information has been carried out in manual ways, and has been considered very time consuming and costly. To overcome the ineffectiveness, this paper introduces a novel framework that can help decision makers discover ideas having the potential to be used in an NPD process. To achieve this, a semi-automatic text mining technique that retrieves useful text patterns from ideas posted on crowdsourcing application is proposed. Then, we provide an online learning algorithm to evaluate whether the idea is potential or not. Finally to verify the effectiveness of our algorithm, we conducted experiments on the data, which are collected from an existing crowd sourcing website.