http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
Deep Learning-based Multiple Pedestrians Detection-Tracking Framework
Xuan-Phung Huynh,Yong-Guk Kim 한국HCI학회 2016 한국HCI학회 학술대회 Vol.2016 No.1
We propose a new Detection-Tracking (DT) framework whereby one can detect a pedestrian, or multiple ones, within a video image and then track them concurrently in a flexible manner. For the detection, a faster R-CNN will be used since it has a state-of-the-art detection accuracy as well as speed. For the tracking, we have developed a fast and reliable tracker, which mainly consists of Kernelized Correlation Filter (KCF) and Kalman filter and shows enhancing performance in the occlusion and human-crossing situations. After the faster R-CNN detects objects’ regions and scores for that objects, our tracker estimates object’s position based on kernel method and Kalman filter. We demonstrate that the proposed framework can detect and track multiple moving pedestrians concurrently for the walking crowd scene.
박상민(Sang-Min Park),Huynh Xuan Phung,김용국(Yong-Guk Kim) 한국HCI학회 2017 한국HCI학회 학술대회 Vol.2017 No.2
본 논문은 야간주행 중 운전자의 시선 방향을 Deep Neural Network(DNN)을 사용하여 검출하는 시스템을 제안한다. 운전자의 시선 방향을 검출하기 위해 선글라스/안경착용 운전자와 착용하지 않은 운전자 각각 차량 내 시선 방향을 8 가지, 눈을 감은 상태 1 가지, 알 수 없는 상태 1 가지, 총 20 가지 시선영역으로 분류한다. 적외선 카메라 (IR)로 촬영한 운전자의 얼굴 영상을 사용하여 DNN 학습 및 실험을 진행하였으며 평균 91% 정확도로 시선영역을 검출할 수 있음을 보였다.
Deep Learning based smile detection for mobile application
Minh Bao Nguyen-Khoa,Phung Xuan Huynh,Yong-Guk Kim 한국HCI학회 2018 한국HCI학회 학술대회 Vol.2018 No.1
Human-computer interaction has become an attractive topic in both computer science and others. People are trying to improve the user interface design as well as the user experience to make things easier for users. Meanwhile, in computer science, many studies pay attention to detecting the facial expressions to help computers efficiently interact with users based on their emotions. In this paper, we propose a convolutional neural network model to detect people’s expression efficiently by using their facial images. The facial expression recognition has been around so long and yet most of the current approaches use hand-crafted features. In this study, we would like to propose a different approach, which is to leverage the feature learning power of convolutional neural networks to handle the task. For validation, the proposed model can reach the average accuracy of 94.2% on the GENKI4K database. Moreover, the model is kept simple and used to develop a demonstrative mobile camera application, which can automatically detect smiles and capture images in real-time.