Chin-Hung Teng
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Chin-Hung Teng has a total of 22 co-authors. Here are the top co-authors by number of shared publications:
- I-Shyan Hwang
Department of Computer Science and Engineering, Yuan Ze University, Taoyuan 32003, Taiwan
2 shared publications AliAkbar Nikoukar
1 shared publication- Andrew Tanny Liem
Department of Computer Science, Universitas Klabat, Manado, Indonesia
1 shared publication Ben-Jian Dong
School of Artificial Intelligence and Electrical Engineering Guizhou Institute of Technology Guizhou China
1 shared publication- Chen-Yuan Hsieh
Department of Information Communication, Yuan Ze University, Chung-Li, Taiwan
1 shared publication Chia-Ming Cheng
Department of Computer Science, National Tsing Hua University, Hsinchu 300, Taiwan
1 shared publication
Latest Publications
- Open Access
- 28 May 2026
- Published by MDPI AG in Journal of Imaging
- Vol. 12 (6) , 233
Brain Tumor Classification in MRI Images Using Combined Transfer Learning and Convolutional Neural Networks
Early and accurate brain tumor detection is vital for effective treatment. We propose a deep learning framework for MRI-based brain tumor classification, featuring a novel Custom CNN evaluated independently alongside six pre-trained models for comparative analysis (InceptionV3, EfficientNetV2L, ResNet152V2, Xception, VGG16, and MobileNetV2). Additionally, three separate ensemble models are constructed to analyze whether model combination improves performance. Experiments conducted on the Kaggle-Multiclass brain MRI dataset show that the proposed Custom CNN achieves the best performance, with an accuracy of 99.54%, and features a task-specific architecture (0.57M parameters) that achieves superior performance through domain-specific feature learning and computational efficiency, thus outperforming both individual pre-trained models and ensemble approaches. Among pre-trained models, EfficientNetV2L (99.47%) and InceptionV3 (99.39%) show competitive results, while the best ensemble model achieves 99.47% accuracy, indicating clinical deployment potential pending external validation. These results demonstrate that the proposed Custom CNN provides superior performance without requiring ensemble complexity, thus highlighting its effectiveness and efficiency for automated brain tumor classification.
- 28 May 2026
- Published by MDPI AG in Journal of Imaging
- Vol. 12 (6) , 233
Open AccessBrain Tumor Classification in MRI Images Using Combined Transfer Learning and Convolutional Neural Networks
Early and accurate brain tumor detection is vital for effective treatment. We propose a deep learning framework for MRI-based brain tumor classification, featuring a novel Custom CNN evaluated independently alongside six pre-trained models for comparative analysis (InceptionV3, EfficientNetV2L, ResNet152V2, Xception, VGG16, and MobileNetV2). Additionally, three separate ensemble models are constructed to analyze whether model combination improves performance. Experiments conducted on the Kaggle-Multiclass brain MRI dataset show that the proposed Custom CNN achieves the best performance, with an accuracy of 99.54%, and features a task-specific architecture (0.57M parameters) that achieves superior performance through domain-specific feature learning and computational efficiency, thus outperforming both individual pre-trained models and ensemble approaches. Among pre-trained models, EfficientNetV2L (99.47%) and InceptionV3 (99.39%) show competitive results, while the best ensemble model achieves 99.47% accuracy, indicating clinical deployment potential pending external validation. These results demonstrate that the proposed Custom CNN provides superior performance without requiring ensemble complexity, thus highlighting its effectiveness and efficiency for automated brain tumor classification.
- Open Access
Guided Image Feature Matching using Feature Spatial Order
Image feature matching plays a vital role in many computer vision tasks. Although many image feature detection and matching techniques have been proposed over the past few decades, it is still time-consuming to match feature points in two images, especially for images with a large number of detected features. Feature spatial order can estimate the probability that a pair of features is correct. Since it is a completely independent concept from epipolar geometry, it can be used to complement epipolar geometry in guiding feature match in a target region so as to improve matching efficiency. In this paper, we integrate the concept of feature spatial order into a progressive matching framework. We use some of the initially matched features to build a computational model of feature spatial order and employs it to calculates the possible spatial range of subsequent feature matches, thus filtering out unnecessary feature matches. We also integrate it with epipolar geometry to further improve matching efficiency and accuracy. Since the spatial order of feature points is affected by image rotation, we propose a suitable image alignment method from the fundamental matrix of epipolar geometry to remove the effect of image rotation. To verify the feasibility of the proposed method, we conduct a series of experiments, including a standard benchmark dataset, self-generated simulated images, and real images. The results demonstrate that our proposed method is significantly more efficient and has more accurate feature matching than the traditional method.
Open AccessGuided Image Feature Matching using Feature Spatial Order
Image feature matching plays a vital role in many computer vision tasks. Although many image feature detection and matching techniques have been proposed over the past few decades, it is still time-consuming to match feature points in two images, especially for images with a large number of detected features. Feature spatial order can estimate the probability that a pair of features is correct. Since it is a completely independent concept from epipolar geometry, it can be used to complement epipolar geometry in guiding feature match in a target region so as to improve matching efficiency. In this paper, we integrate the concept of feature spatial order into a progressive matching framework. We use some of the initially matched features to build a computational model of feature spatial order and employs it to calculates the possible spatial range of subsequent feature matches, thus filtering out unnecessary feature matches. We also integrate it with epipolar geometry to further improve matching efficiency and accuracy. Since the spatial order of feature points is affected by image rotation, we propose a suitable image alignment method from the fundamental matrix of epipolar geometry to remove the effect of image rotation. To verify the feasibility of the proposed method, we conduct a series of experiments, including a standard benchmark dataset, self-generated simulated images, and real images. The results demonstrate that our proposed method is significantly more efficient and has more accurate feature matching than the traditional method.
- Open Access
- 18 September 2025
- Published by MDPI AG in Applied Sciences
- Vol. 15 (18) , 10161
Exploring User Engagement and Purchase Intentions in T-Shirt Retail Through Augmented Reality and Instagram Filters
Augmented reality (AR) technologies—such as Instagram filters—bridge the digital and physical worlds by allowing users to virtually try on clothing, thereby reducing the risk of virus transmission. In the T-shirt retail industry, AR enables product personalization, decreases the need for physical production, minimizes textile waste, and lowers carbon emissions. It also benefits individuals with limited mobility or those who prefer shopping online. This study tested several hypotheses on 105 active Instagram filter users using filters from the ’Apprecio’ account on mobile devices. Data analyzed using the partial least squares method revealed that interactivity significantly influences both purchase intention and continued use of digital platforms. While hedonic and vivid features enhance the user experience, they have a limited impact on driving purchases or long-term engagement. Customers’ engagement and buying intent are more strongly shaped by practical and interactive elements. The study recommends that companies invest in developing interactive AR features to boost customer satisfaction and foster trust. Future research should involve larger participant samples and investigate specific interactive elements—such as virtual try-on tools—to better understand their impact on consumer behavior. This study highlights the critical role of interactivity in AR for delivering meaningful and engaging shopping experiences.
- 18 September 2025
- Published by MDPI AG in Applied Sciences
- Vol. 15 (18) , 10161
Open AccessExploring User Engagement and Purchase Intentions in T-Shirt Retail Through Augmented Reality and Instagram Filters
Augmented reality (AR) technologies—such as Instagram filters—bridge the digital and physical worlds by allowing users to virtually try on clothing, thereby reducing the risk of virus transmission. In the T-shirt retail industry, AR enables product personalization, decreases the need for physical production, minimizes textile waste, and lowers carbon emissions. It also benefits individuals with limited mobility or those who prefer shopping online. This study tested several hypotheses on 105 active Instagram filter users using filters from the ’Apprecio’ account on mobile devices. Data analyzed using the partial least squares method revealed that interactivity significantly influences both purchase intention and continued use of digital platforms. While hedonic and vivid features enhance the user experience, they have a limited impact on driving purchases or long-term engagement. Customers’ engagement and buying intent are more strongly shaped by practical and interactive elements. The study recommends that companies invest in developing interactive AR features to boost customer satisfaction and foster trust. Future research should involve larger participant samples and investigate specific interactive elements—such as virtual try-on tools—to better understand their impact on consumer behavior. This study highlights the critical role of interactivity in AR for delivering meaningful and engaging shopping experiences.
- Open Access
Enhancing Tactile Internet Reliability: AI-Driven Resilience in NG-EPON Networks
To guarantee the reliability of Tactile Internet (TI) applications such as telesurgery, which demand extremely high reliability and are experiencing rapid expansion, we propose a novel smart resilience mechanism for Next-Generation Ethernet Passive Optical Networks (NG-EPONs). Our architecture integrates Artificial Intelligence (AI) and Software-Defined Networking (SDN)-Enabled Broadband Access (SEBA) platform to proactively enhance network reliability and performance. By harnessing the AI’s capabilities, our system automatically detects and localizes fiber faults, establishing backup communication links using Radio Frequency over Glass (RFoG) to prevent service disruptions. This empowers NG-EPONs to maintain uninterrupted, high-quality network service even in the face of unexpected failures, meeting the stringent Quality-of-Service (QoS) requirements of critical TI applications. Our AI model, rigorously validated through 5-fold cross-validation, boasts an average accuracy of 81.49%, with a precision of 84.33%, recall of 78.18%, and F1-score of 81.00%, demonstrating its robust performance in fault detection and prediction. The AI model triggers immediate corrective actions through the SDN controller. Simulation results confirm the efficacy of our proposed mechanism in terms of delay, system throughputs and packet drop rate, and bandwidth waste, ultimately ensuring the delivery of high-quality network services.
Open AccessEnhancing Tactile Internet Reliability: AI-Driven Resilience in NG-EPON Networks
To guarantee the reliability of Tactile Internet (TI) applications such as telesurgery, which demand extremely high reliability and are experiencing rapid expansion, we propose a novel smart resilience mechanism for Next-Generation Ethernet Passive Optical Networks (NG-EPONs). Our architecture integrates Artificial Intelligence (AI) and Software-Defined Networking (SDN)-Enabled Broadband Access (SEBA) platform to proactively enhance network reliability and performance. By harnessing the AI’s capabilities, our system automatically detects and localizes fiber faults, establishing backup communication links using Radio Frequency over Glass (RFoG) to prevent service disruptions. This empowers NG-EPONs to maintain uninterrupted, high-quality network service even in the face of unexpected failures, meeting the stringent Quality-of-Service (QoS) requirements of critical TI applications. Our AI model, rigorously validated through 5-fold cross-validation, boasts an average accuracy of 81.49%, with a precision of 84.33%, recall of 78.18%, and F1-score of 81.00%, demonstrating its robust performance in fault detection and prediction. The AI model triggers immediate corrective actions through the SDN controller. Simulation results confirm the efficacy of our proposed mechanism in terms of delay, system throughputs and packet drop rate, and bandwidth waste, ultimately ensuring the delivery of high-quality network services.
- 15 June 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 20 (21) , 12980-12991
Modularized Multi-Stage Binarized Deep-Model Compression and Performance Recovery
Despite the approaching Internet of Things (IoT) era, most smart devices cannot directly execute complicated deep neu-ral networks (DNNs) due to their limited memory sizes and com-puting power. Even with model compression, compressing and ex-ecuting a non-binarized DNN model often needs a codebook for weight calculation, which typically requires acceleration from a Graphics Processing Unit (GPU). Parameter binarization is one commonly used method to reduce computing time and space, but it often seriously compromises the accuracy of the binarized DNN (BNN), which tends to adopt single-stage non-modularized design. To address these issues, we propose a modularized multi-stage BNN model compression with flexible combination of parameter and structure compression to minimize model size while maintain-ing maximum accuracy possible. The compressed BNNs do not need any codebooks so they better fit into embedded platforms without any GPU acceleration. From our evaluations, the size of a BNN model can be reduced to 1/57 size at the cost of only about a 5% decrease in accuracy, which outperforms the state-of-the-art single-stage compression on BNN models. Furthermore, the pro-posed method can recover the accuracy of a BNN by 5% even after its size has been further compressed to approximately half its size, demonstrating its potentials in leveraging edge intelligence.
- 15 June 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 20 (21) , 12980-12991
Modularized Multi-Stage Binarized Deep-Model Compression and Performance Recovery
Despite the approaching Internet of Things (IoT) era, most smart devices cannot directly execute complicated deep neu-ral networks (DNNs) due to their limited memory sizes and com-puting power. Even with model compression, compressing and ex-ecuting a non-binarized DNN model often needs a codebook for weight calculation, which typically requires acceleration from a Graphics Processing Unit (GPU). Parameter binarization is one commonly used method to reduce computing time and space, but it often seriously compromises the accuracy of the binarized DNN (BNN), which tends to adopt single-stage non-modularized design. To address these issues, we propose a modularized multi-stage BNN model compression with flexible combination of parameter and structure compression to minimize model size while maintain-ing maximum accuracy possible. The compressed BNNs do not need any codebooks so they better fit into embedded platforms without any GPU acceleration. From our evaluations, the size of a BNN model can be reduced to 1/57 size at the cost of only about a 5% decrease in accuracy, which outperforms the state-of-the-art single-stage compression on BNN models. Furthermore, the pro-posed method can recover the accuracy of a BNN by 5% even after its size has been further compressed to approximately half its size, demonstrating its potentials in leveraging edge intelligence.
- 9 August 2017
- Published by Springer Nature in The Visual Computer
- Vol. 34 (11) , 1507-1523
Reconstructing three-dimensional models of objects using a Kinect sensor
- 9 August 2017
- Published by Springer Nature in The Visual Computer
- Vol. 34 (11) , 1507-1523
Reconstructing three-dimensional models of objects using a Kinect sensor
- 21 April 2017
- Published by SAGE Publications in Journal of Educational Computing Research
- Vol. 56 (2) , 254-271
Impact of Augmented Reality on Programming Language Learning: Efficiency and Perception
Although the learning of programming language is critical in science and technology education, it might be difficult for some students, especially novices. One possible reason might be the fact that programming language, especially for three-dimensional (3D) applications, is too complex and abstract for these students to understand. Programming for 3D applications requires understanding the spatial relationship of 3D objects and hence needs a visualization technique more. In view of this, this article presents an augmented reality (AR)-enhanced learning system that offers visual representation and interactivity to help students learn programming for 3D applications. To examine the influences of such an AR-enhanced system on student learning, a within-group experiment with 34 college students was conducted. All students used both of an AR-enhanced version and an ordinary version. The findings revealed that the AR-enhanced version made students have better learning efficiency than the ordinary system. In addition, the AR-enhanced system also made students have enhanced perceptions in terms of system usability, flow experience, and usage perception. Based on the results, further development of AR-enhanced learning systems is also suggested and discussed.
- 21 April 2017
- Published by SAGE Publications in Journal of Educational Computing Research
- Vol. 56 (2) , 254-271
Impact of Augmented Reality on Programming Language Learning: Efficiency and Perception
Although the learning of programming language is critical in science and technology education, it might be difficult for some students, especially novices. One possible reason might be the fact that programming language, especially for three-dimensional (3D) applications, is too complex and abstract for these students to understand. Programming for 3D applications requires understanding the spatial relationship of 3D objects and hence needs a visualization technique more. In view of this, this article presents an augmented reality (AR)-enhanced learning system that offers visual representation and interactivity to help students learn programming for 3D applications. To examine the influences of such an AR-enhanced system on student learning, a within-group experiment with 34 college students was conducted. All students used both of an AR-enhanced version and an ordinary version. The findings revealed that the AR-enhanced version made students have better learning efficiency than the ordinary system. In addition, the AR-enhanced system also made students have enhanced perceptions in terms of system usability, flow experience, and usage perception. Based on the results, further development of AR-enhanced learning systems is also suggested and discussed.
- 16 March 2017
- Published by Springer Nature
- p. 136-147
Image-Based Camera Localization for Large and Outdoor Environments
- 16 March 2017
- Published by Springer Nature
- p. 136-147
Image-Based Camera Localization for Large and Outdoor Environments
- 28 November 2013
- Published by Springer Nature in Machine Vision and Applications
- Vol. 25 (2) , 421-436
Improving three-dimensional point reconstruction from image correspondences using surface curvatures
- 28 November 2013
- Published by Springer Nature in Machine Vision and Applications
- Vol. 25 (2) , 421-436
Improving three-dimensional point reconstruction from image correspondences using surface curvatures
- 11 March 2013
- Published by Optica Publishing Group in Journal of Optical Communications and Networking
- Vol. 5 (4) , 271-282
Scalable Architecture for VOD Service Enhancement Based on a Cache Scheme in an Ethernet Passive Optical Network
Optics InfoBase is the Optical Society's online library for flagship journals, partnered and copublished journals, and recent proceedings from OSA conferences.
- 11 March 2013
- Published by Optica Publishing Group in Journal of Optical Communications and Networking
- Vol. 5 (4) , 271-282
Scalable Architecture for VOD Service Enhancement Based on a Cache Scheme in an Ethernet Passive Optical Network
Optics InfoBase is the Optical Society's online library for flagship journals, partnered and copublished journals, and recent proceedings from OSA conferences.