|
Kyomin Hwang I am a MS student in Seoul National University, studying deep learning and computer vision under the supervision of professor Nojun Kwak. Previously, I did my undergrad at Kookmin University (Computer Science major). Email / CV / Github / Google Scholar |
|
|
|
|
Kyomin Hwang*, Gangjoon Yoon*, Jinjoo Song, Sang Min Yoon Engineering Applications of Artificial Intelligence sciencedirect Image super resolution, which obtains high resolution output from a corresponding low resolution image, has been challenging due to inefficiencies in establishing complex high dimensional mapping for massive raw data. Single image super resolution can dramatically improve performance compared with current algorithms due to the proliferation of deep learning systems. However, convolutional kernels in deep neural networks are locally connected to the input feature maps, whereas features only interact with their local neighbors. Mutual interference between local features without considering global features causes blurring and staircase effects. This paper proposes an end-to-end single image super resolution model by simultaneously separating high and low frequency features and learning adaptive local and global features to effectively reconstruct the high resolution image by minimizing the loss of edges and texture information. The image frequency decomposition module with an attention block emphasizes self-representative low frequency features to separate high and low frequency features. The bidirectional global and local feature exchange module extracts global and local features from the separated network and fuses each feature to improve performance. Quantitative and qualitative analyses for the proposed frequency adaptive network validated that the proposed method is stable and robust against blurring and staircase effects by separating texture and the structure into adaptive and shared networks |
|
|
|
Kyomin Hwang, Sang-Chul Kim 2021 International Conference on Information Networking (ICOIN) IEEE In this paper, we studied a model that can distinguish several different human behaviors. We trained CSI data using the Convolutional Neural Network algorithm. The suggested model showed 94.597% accuracy in distinguishing seven different human activities. |
|
|
|
Kyomin Hwang*, Suyoung Kim*, Junhoo Lee*, Nojun Kwak 2024 CVPRW(Responsible Generative AI Workshop) Large Models (LMs) have heightened expectations for the potential of general AI as they are akin to human in telligence. This paper shows that recent large models such as Stable Diffusion and DALL-E3 also share the vulnera bility of human intelligence, namely the “white bear phe nomenon”. We investigate the causes of the white bear phe nomenon by analyzing their representation space. Based on this analysis, we propose a simple prompt-based at tack method, which generates figures prohibited by the LM provider’s policy. To counter these attacks, we introduce prompt-based defense strategies inspired by cognitive ther apy techniques, successfully mitigating attacks by up to 48.22%. |
|
2024 CVPRW Emerging of foundation model on deep learning fields such as computer vision and natural language processing, finetuning for various downstream tasks has become more important. In this paper, we present a methodology for efficient finetuning for the pre-trained model when a small target dataset is given. In order to learn new domains while maintaining the knowledge of the pre-trained model, the idea of domain adaptation is adopted to pursue continuous and efficient finetuning. Our work is performed for the image captioning task, a dataset provided by CVPR NICE workshop 2023. Using BLIP-2 as the baseline model, the concepts of adapter with zero convolution and EMA are added along with finetuning layer normalization. To avoid overfitting to the target dataset during the training process, the batch is constructed by mixing with the CC3M dataset. Achieving outstanding results in the workshops by using this methodology, we demonstrate the emerging capability of this finetuning method. |
|
|
|
Inseop Chung, Kyomin Hwang, Jayeon Yoo, Nojun Kwak Arxiv Arxiv This paper mitigates this issue to improve performance in the CTA scenario. To alleviate the bias issue, we make class-wise exponential moving average target prototypes with reliable target samples and exploit them to cluster the target features class-wisely. Moreover, we aim to align the target distributions to the source distribution by anchoring the target feature to its corresponding source prototype. |
|
Junhoo Lee, Hyunho Lee, Kyomin Hwang, Nojun Kwak Arxiv Arxiv While the success of deep learning is commonly attributed to its theoretical equivalence with Support Vector Machines (SVM), the practical implications of this relationship have not been thoroughly explored. This paper pioneers an exploration in this domain, specifically focusing on the identification of Deep Support Vectors (DSVs) within deep learning models. We introduce the concept of DeepKKT conditions, an adaptation of the traditional Karush-Kuhn-Tucker (KKT) conditions tailored for deep learning. Through empirical investigations, we illustrate that DSVs exhibit similarities to support vectors in SVM, offering a tangible method to interpret the decisionmaking criteria of models. Additionally, our findings demonstrate that models can be effectively reconstructed using DSVs, resembling the process in SVM. |
|
|
|
황교민, 장민혁, 이윤재 CEIC 2021 전자정보통신학술대회 We propose a LSTM model with feature attention module for Industrial Control System Threat Detection using HAI dataset. |
|
|
김상철, 황교민 한국통신학회 학술대회논문집, 855-856 |
|
|
|
Code Video Recently, IoT technology is also being used in care services. Representative examples include KT's IoT-based location tracker, safe LED solution, and monitoring system through TV viewing. However, all of these are not services that monitor the mental health of the elderly living alone, but mainly focus on physical health monitoring through location identification of the elderly living alone. In other words, The use of IoT technology has enabled many elderly people living alone to enjoy benefits, but it is difficult to detect the most important early because they do not focus on emotional and mental health monitoring. Depression is one of the most well-treated psychiatric diseases if detected early, so the majority of people can recover to their normal daily lives with quick detection and proper treatment. Therefore, We introduce services that monitor the emotional state of single-person households (especially the elderly living alone) to help detect and treat depression early. |
|
|
The template is from Jon Barron. Thank you for sharing! |