Shih-Chia Huang

Prof. Dr. Shih-Chia Huang

Department of Electronic Engineering, National Taipei University of Technology NTUT Office 405, No.1, Sec. 3, Chung Hsiao E. Rd, Taipei 106, Taiwan

Research Focus & Metrics

Academic Career Length

15Years (2012-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
22 October 2012 in IEEE Transactions on Image Processing
Verified
Journal Article
Closed access
Shih-Chia HuangShih-Chia Huang
Shih-Chia Huang
FCFan-Chieh Cheng
Fan-Chieh Cheng
YCYi-Sheng Chiu
Yi-Sheng Chiu
This paper proposes an efficient method to modify histograms and enhance contrast in digital images. Enhancement plays a significant role in digital image processing, computer vision, and pattern recognition. We present an automatic transformation technique that improves the brightness of dimmed images via the gamma correction and probability distribution of luminance pixels. To enhance video, the proposed image-enhancement method uses temporal information regarding the differences between each frame to reduce computational complexity. Experimental results demonstrate that the proposed method produces enhanced images of comparable or higher quality than those produced using previous state-of-the-art methods.

3 March 2020 in IEEE Transactions on Pattern Analysis and Machine Intelligence
Verified
Journal Article
Closed access
Shih-Chia HuangShih-Chia Huang
Shih-Chia Huang
TLTrung-Hieu Le
Trung-Hieu Le
DJDa-Wei Jaw
Da-Wei Jaw
In the past half of the decade, object detection approaches based on convolutional neural network have been widely studied and successfully applied in many computer vision applications. However, detecting objects in inclement weather conditions remains a major challenge because of poor visibility. In this paper, we address the object detection problem in the presence of fog by introducing a novel dual-subnet network (DSNet) that can be trained end-to-end and jointly learn three tasks: visibility enhancement, object classification, and object localization. DSNet attains complete performance improvement by including two subnetworks: detection subnet and restoration subnet. We employ RetinaNet as a backbone network (also called detection subnet), which is responsible for learning to classify and locate objects. The restoration subnet is designed by sharing feature extraction layers with the detection subnet and adopting a feature recovery (FR) module for visibility enhancement. Experimental results show that our DSNet achieved 50.84% mean average precision (mAP) on a synthetic foggy dataset that we composed and 41.91% mAP on a public natural foggy dataset (Foggy Driving dataset), outperforming many state-of-the-art object detectors and combination models between dehazing and detection methods while maintaining a high speed.

7 May 2018 in IEEE Journal of Biomedical and Health Informatics
Verified
Journal Article
Closed access
MGMunkhjargal Gochoo
Munkhjargal Gochoo
Tan-Hsu TanTan-Hsu Tan
Tan-Hsu Tan
Shing-Hong LiuShing-Hong Liu
Shing-Hong Liu
FJFu-Rong Jean
Fu-Rong Jean
FAFady Alnajjar
Fady Alnajjar
Shih-Chia HuangShih-Chia Huang
Shih-Chia Huang
Elderly population (over the age of 60) is predicted to be 1.2 billion by 2025. Most of the elderly people would like to stay alone in their own house due to the high eldercare cost and privacy invasion. Unobtrusive activity recognition is the most preferred solution for monitoring daily activities of the elderly people living alone rather than the camera and wearable devices based systems. Thus, we propose an unobtrusive activity recognition classifier using Deep Convolutional Neural Network (DCNN) and anonymous binary sensors (PIR sensors and door sensors). The real world long-term fully annotated Aruba open dataset collected by binary sensors has been employed for the training and evaluation of the classifier. First, ten basic daily activities, namely: Bed_to_Toilet, Eating, Meal_Preparation, Relax, Sleeping, Work, Housekeeping, Wash_Dishes, Enter_Home, and Leave_Home are segmented with different sliding window sizes, and then converted into binary activity images. Next, the activity images are employed for training the DCNN classifiers with different parameters. Finally, the trained classifiers are evaluated with the 10-fold cross validation method, and the results show that the best DCNN classifier gives 0.79 and 0.951 of F1-score for all 10 activities and eight activities (excluding Leave_Home and Wash_Dishes), respectively.

6 May 2019 in IEEE Internet of Things Journal
Verified
Journal Article
Closed access
MGMunkhjargal Gochoo
Munkhjargal Gochoo
TTTan-Hsu Tan
Tan-Hsu Tan
Shih-Chia HuangShih-Chia Huang
Shih-Chia Huang
TBTsedevdorj Batjargal
Tsedevdorj Batjargal
JHJun-Wei Hsieh
Jun-Wei Hsieh
Yung-Fu ChenYung-Fu Chen
Yung-Fu Chen
In recent years, the number of yoga practitioners has been drastically increased and there are more men and older people practice yoga than ever before. IoT-based yoga training system is needed for those who want to practice yoga at home. Some studies have proposed RGB/Kinect camera-based or wearable device-based yoga posture recognition methods with a high accuracy; however, the former has a privacy issue and the latter is impractical in the long-term application. Thus, this study proposes an IoT-based privacy-preserving yoga posture recognition system employing a deep convolutional neural network (DCNN) and a low-resolution infrared sensor-based wireless sensor network (WSN). The WSN has three nodes (x, y, and z-axes) where each integrates 8×8 pixels’ thermal sensor module and a Wi-Fi module for connecting the deep learning server. We invited 18 volunteers to perform 26 yoga postures for two sessions each lasted for 20 seconds. First, recorded sessions are saved as.csv files, then preprocessed and converted to grayscale posture images. Totally, 93200 posture images are employed for the validation of the proposed DCNN models. The 10-fold cross-validation results revealed that F1-scores of the models trained with xyz (all 3-axes) and y (only y-axis) posture images were 0.9989 and 0.9854, respectively. An average latency for a single posture image classification on the server was 107 ms. Thus, we conclude that the proposed IoT-based yoga posture recognition system has a great potential in the privacy-preserving yoga training system.

17 June 2020 in IEEE Transactions on Circuits and Systems for Video Technology
Verified
Journal Article
Closed access
DJDa-Wei Jaw
Da-Wei Jaw
Shih-Chia HuangShih-Chia Huang
Shih-Chia Huang
SKSy-Yen Kuo
Sy-Yen Kuo
In this paper, we present a simple, efficient, and highly modularized network architecture for single-image snowremoval. To address the challenging snow-removal problem in terms of network interpretability and computational complexity, we employ a pyramidal hierarchical design with lateral connections across different resolutions. This design enables us to incorporate high-level semantic features with other feature maps at different scales to enrich location information and reduce computational time. In addition, a refinement stage based on recently introduced generative adversarial networks (GANs) is proposed to further improve the visual quality of the resulting snow-removed images and make a refined image and a clean image indistinguishable by a computer vision algorithm to avoid the potential perturbations of machine interpretation. Finally, atrous spatial pyramid pooling (ASPP) is adopted to probe features at multiple scales and further boost the performance. The proposed DesnowGAN (DS-GAN) performs significantly better than state-of-the-art methods quantitatively and qualitatively on the Snow100K dataset.

23 August 2018 in IEEE Sensors Journal
Verified
Journal Article
Closed access
TTTan-Hsu Tan
Tan-Hsu Tan
MGMunkhjargal Gochoo
Munkhjargal Gochoo
Shih-Chia HuangShih-Chia Huang
Shih-Chia Huang
YLYi-Hung Liu
Yi-Hung Liu
Shing-Hong LiuShing-Hong Liu
Shing-Hong Liu
Yung-Fa HuangYung-Fa Huang
Yung-Fa Huang
In the last decade, unobtrusive (device-free and non-privacy invasive) recognition of activities of daily living (ADLs) for an individual in a smart home has been studied by many researchers. However, the unobtrusive recognition of multi-resident activities in a smart home is hardly studied. We propose a novel RGB activity image-based DCNN classifier for the unobtrusive recognition of the multi-resident activities (Bed_to_Toilet, Bed, Breakfast, Lunch, Leave_home, Laundry, Dinner, Night_wandering, R2_work, and R1_medicine) using Cairo open dataset provided by CASAS project. The open dataset is collected by environmental sensors (PIR and temperature sensors) in Cairo testbed while an adult couple with a dog was living for 55 days. The dataset is preprocessed with activity segmentation, sliding window, and RGB activity image conversion steps. The experimental results demonstrate that our classifier has the highest total accuracy of 95.2% among the previously developed machine learning classifiers that employed the same dataset. Moreover, the proposed RGB activity image was proven to be helpful for increasing the recognition rate. Therefore, we conclude that the proposed DCNN classifier is a useful tool for the unobtrusive recognition of the multi-resident activity in a home.