[CSE7001/8001] Creative Software (2026 Fall)

Date:  09/04 (offline, B09)

Speaker: Dr. Sung-Ho Bae, Associate Professor, Kyung Hee University

Title: Deep Neural Networks and Their Applications 

Abstract: This presentation provides an basic overview of deep neural networks and their applications. Especially, this presentation covers basic architectures and types of deep neural networks depending on learning strategies. Also, this presentation aims to explan the fundamental mechanism of deep neural networks in relation of human brain systems. 

Date: 09/18 (offline, B09)

Speaker: Dr. Chaeyong Park,  Assistant Professor, Korea University

Title: Enhancing Immersive VR Interaction through Haptics and HCI Technologies

Abstract: Achieving realistic and immersive interactions in virtual reality (VR) requires the seamless integration of multimodal feedback, including visual, auditory, and haptic cues. Among these modalities, haptics plays a particularly important role by providing physical sensations across different parts of the body, from the hands and arms to the torso, through feedback such as vibration, impact, and temperature. This seminar presents an overview of haptic technologies, human-computer interaction (HCI) techniques, and interaction methods for virtual environments. It will introduce key research perspectives, recent advances, and practical approaches to designing haptic interactions. In addition, the seminar will highlight several research projects conducted by the speaker, illustrating how different haptic feedback techniques can be designed and applied to enhance presence, realism, and user experience in immersive VR.

Date: 11/13 (online)

Speaker: Dr. Joonhyung Park,  Assistant Professor, Kyung Hee University

Title: Perception and Reasoning Like Humans and Beyond

Abstract: This talk presents a research journey toward building AI systems that perceive and reason in more human-like ways - and eventually beyond human capabilities. It will highlight work ranging from structured representations for multimodal understanding to reasoning in multimodal LLMs, including research that has been applied in Amazon Quick Automate. The talk will then focus on recent work in test-time scaling for autoregressive image generation, exploring how scaling inference-time computation can elicit the model’s full potential without additional training. Overall, the talk connects these efforts through a common goal: developing AI models that can better understand structure, reason reliably, and make more effective decisions.

Date: 11/20 (online)

Speaker: Dr. Sungahn Ko,  Associate Professor, POSTECHDr. Sungahn Ko, Associate Professor, POSTECH 

Title: Toward Human-Agent Interaction with Visual Interfaces 

Abstract: Recent advances in artificial intelligence (AI) have created new opportunities for integrating AI into everyday work. However, enabling effective human–AI collaboration remains challenging. While users expect timely, interactive assistance from AI systems, meeting these expectations requires both a deep understanding of user tasks and effective communication between humans and AI. In this context, visual interfaces play a critical role by serving as the primary medium for interaction, interpretation, and collaboration. In this talk, I will present how visual interfaces can foster effective human–AI collaboration through two case studies from different application domains. The first case study introduces a visual analytics system that enables traffic congestion experts to analyze road conditions and make broadcasting decisions with the assistance of a deep learning model. The second case study presents a novel AI-assisted visual interface that helps novice users design mobile user interfaces by providing real-time design suggestions from large AI models. Experimental results show that interfaces created with our system receive significantly higher quality scores than those created without AI assistance. In addition, participants reported that the proposed interface was more enjoyable, effective, efficient, and comfortable to use than conventional design tools. Finally, I will introduce our recent work on an SNS-style multi-agent visual analytics system and an empirical study investigating how users intervene during collaboration with web-based AI agents. Together, these examples illustrate how carefully designed visual interfaces can bridge the gap between human expertise and AI capabilities, enabling more effective, transparent, and trustworthy human–AI collaboration.

Date: 11/27 (TBD)

Speaker: Dr.Jihyong Oh, Assistant Professor, CAU

Title: Generative Visual Al from Pixels to 4D Worlds

Abstract: Recent advances in generative AI are rapidly reshaping visual intelligence, from pixel-level restoration and enhancement to image and video generation, editing, and dynamic 3D/4D world modeling. In this talk, I will provide an overview of recent research trends in generative visual AI, highlighting how generative priors, multimodal models, and geometric representations are transforming the way visual content is created, restored, and understood. I will also introduce selected examples from our recent research and discuss emerging directions toward agentic visual systems and generative world models.