김동현 교수

Learning transferable and domain adaptive feature representations from videos is important for video-relevant tasks such as action recognition. Existing video domain adaptation methods mainly rely on adversarial feature alignment, which has been derived from the RGB image space. However, video data is usually associated with multi-modal information, e.g., RGB and optical flow, and thus it remains a challenge to design a better method that considers the cross-modal inputs under the cross-domain a

In this study, we proposed a data-driven approach to the condition monitoring of the marine engine. Although several unsupervised methods in the maritime industry have existed, the common limitation was the interpretation of the anomaly; they do not explain why the model classifies specific data instances as an anomaly. This study combines explainable AI techniques with anomaly detection algorithm to overcome the limitation above. As an explainable AI method, this study adopts Shapley Additive e

An 81.6 GOPS object recognition processor is developed by using NoC and visual image processing (VIP) memory. SIFT (scale invariant feature transform) object recognition requires huge computing power and data transactions among tasks. The chip integrates 10 SIMD PEs for data/task level parallelism while the NoC facilitates inter-PE communications. The VIP memory searches local maximum pixel inside a 3times3 window in a single cycle providing 65.6 GOPS. The proposed processor achieves 15.9 fps SI

Existing vision-language methods typically support two languages at a time at most. In this paper, we present a modular approach which can easily be incorporated into existing vision-language methods in order to support many languages. We accomplish this by learning a single shared Multimodal Universal Language Embedding (MULE) which has been visually-semantically aligned across all languages. Then we learn to relate MULE to visual data as if it were a single language. Our method is not architec

This paper presents the algorithm and implementation of a new high-performance functional unit for floating-point four-dimensional vector inner product (4D dot product; DP4), which is most frequently performed in 3D graphics application. The proposed IEEE-compliant DP4 unit computes Z = AB + CD + EF + GH in one path and keeps the intermediate rounding by IEEE-754 rounding to nearest even. The intermediate rounding is merged with shift alignment, and intermediate carry-propagated addition and nor

Existing unsupervised domain adaptation methods aim to transfer knowledge from a label-rich source domain to an unlabeled target domain. However, obtaining labels for some source domains may be very expensive, making complete labeling as used in prior work impractical. In this work, we investigate a new domain adaptation scenario with sparsely labeled source data, where only a few examples in the source domain have been labeled, while the target domain is unlabeled. We show that when labeled sou

This study proposes an unsupervised anomaly detection method using sensor streams from the marine engine to detect the anomalous system behavior, which may be a possible sign of system failure. Previous works on marine engine anomaly detection proposed a clustering-based or statistical control chart-based approach that is unstable according to the choice of hyperparameters, or cannot fit well to the high-dimensional dataset. As a remedy to this limitation, this study adopts an ensemble-based app

Many self-supervised learning (SSL) methods have been successful in learning semantically meaningful visual representations by solving pretext tasks. However, prior work in SSL focuses on tasks like object recognition or detection, which aim to learn object shapes and assume that the features should be invariant to concepts like colors and textures. Thus, these SSL methods perform poorly on downstream tasks where these concepts provide critical information. In this paper, we present an SSL frame

Optical flow is a vision-based approach that is used for tracking the movement of objects. This robust technique can be an effective tool for determining the source of failures on slope surfaces, including the dynamic behavior of rockfall. However, optical flow-based measurement still remains an issue as the data from optical flow algorithms can be affected by the varied photographing environment, such as weather and illuminations. To address such problems, this paper presents an optical flow-ba

Against steganography to hide secret messages into an innocent-like cover, steganalysis was studied to detect the presence of hidden messages, and steganography flaws were determined by human intervention. In this paper, we present a steganalysis method using deep learning for spatial domain steganography which does not require human intervention. The deep learning-based steganalysis model is designed to have 1 high pass filter, 2 convolutional layers and 2 full connected layers. After being tra

주거환경정비사업은 사업기간이 길고, 목적을 달리하는 이해관계자가 많아 추진과정에서 많은 문제점을 낳는다. 본 연구에서는 주거환경정비사업의 전 과정에서 사업지연을 초래하는 요소를 도출하고, 주거환경정비사업의 원활한 진행을 평가할 수 있는 지표를 설정하였다. 이를 통하여 주거환경정비사업 대상지의 입지선정 이후 효율적ㆍ합리적인 주거환경정비사업을 위한 관리시스템을 구축하는데 기초자료를 제시하고 정책방향을 제언하였다. 연구방법으로는 먼저 사업지연요소를 도출하기 위해 문헌조사, 법ㆍ제도 검토, 언론기사 분석을 실시하였으며, 지표 유형화를 위해 전문가 심층면접, FGI(Focus Group Interview), 예비설문조사를 진행하였다. 이렇게 완성된 최종지표를 가지고 AHP기법을 활용하여 관련분야(조합, 시행ㆍ부동산ㆍ감정평가, 설계ㆍ시공, 학계ㆍ연구원, 공무원)의 전문가를 대상으로 설문조사를 실시하였다. 설문결과를 바탕으로 사업지연에 영향을 미치는 요소별 중요도 및 관련분야의 집단별 중요도를