Chun-Fu Chen
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
Chun-Fu Chen has a total of 52 co-authors. Here are the top co-authors by number of shared publications:
- Rameswar Panda
MIT-IBM Watson AI Lab
2 shared publications Anthony F. Jacobine
Henkel Corporation USA
1 shared publication- Ashraful Islam
Rensselaer Polytechnic Institute New York USA
1 shared publication - Bernd Mayer
Fraunhofer Institute for Manufacturing Technology and Advanced Materials IFAM, 28359 Bremen, Germany
1 shared publication Bernhard Schöttmer
Henkel AG & Co. KGaA Germany
1 shared publicationBrendan Kneafsey
Loctite (R&D) Limited Ireland
1 shared publication
Latest Publications
- 27 March 2023
- Published by Wiley
- p. 1-29
Structural Epoxy Adhesives
- Open Access
RegionViT: Regional-to-Local Attention for Vision Transformers
Vision transformer (ViT) has recently shown its strong capability in achieving comparable results to convolutional neural networks (CNNs) on image classification. However, vanilla ViT simply inherits the same architecture from the natural language processing directly, which is often not optimized for vision applications. Motivated by this, in this paper, we propose a new architecture that adopts the pyramid structure and employ a novel regional-to-local attention rather than global self-attention in vision transformers. More specifically, our model first generates regional tokens and local tokens from an image with different patch sizes, where each regional token is associated with a set of local tokens based on the spatial location. The regional-to-local attention includes two steps: first, the regional self-attention extract global information among all regional tokens and then the local self-attention exchanges the information among one regional token and the associated local tokens via self-attention. Therefore, even though local self-attention confines the scope in a local region but it can still receive global information. Extensive experiments on four vision tasks, including image classification, object and keypoint detection, semantics segmentation and action recognition, show that our approach outperforms or is on par with state-of-the-art ViT variants including many concurrent works. Our source codes and models are available at https://github.com/ibm/regionvit.
Open AccessRegionViT: Regional-to-Local Attention for Vision Transformers
Vision transformer (ViT) has recently shown its strong capability in achieving comparable results to convolutional neural networks (CNNs) on image classification. However, vanilla ViT simply inherits the same architecture from the natural language processing directly, which is often not optimized for vision applications. Motivated by this, in this paper, we propose a new architecture that adopts the pyramid structure and employ a novel regional-to-local attention rather than global self-attention in vision transformers. More specifically, our model first generates regional tokens and local tokens from an image with different patch sizes, where each regional token is associated with a set of local tokens based on the spatial location. The regional-to-local attention includes two steps: first, the regional self-attention extract global information among all regional tokens and then the local self-attention exchanges the information among one regional token and the associated local tokens via self-attention. Therefore, even though local self-attention confines the scope in a local region but it can still receive global information. Extensive experiments on four vision tasks, including image classification, object and keypoint detection, semantics segmentation and action recognition, show that our approach outperforms or is on par with state-of-the-art ViT variants including many concurrent works. Our source codes and models are available at https://github.com/ibm/regionvit.
- Open Access
Dynamic Distillation Network for Cross-Domain Few-Shot Recognition with Unlabeled Data
Most existing works in few-shot learning rely on meta-learning the network on a large base dataset which is typically from the same domain as the target dataset. We tackle the problem of cross-domain few-shot learning where there is a large shift between the base and target domain. The problem of cross-domain few-shot recognition with unlabeled target data is largely unaddressed in the literature. STARTUP was the first method that tackles this problem using self-training. However, it uses a fixed teacher pretrained on a labeled base dataset to create soft labels for the unlabeled target samples. As the base dataset and unlabeled dataset are from different domains, projecting the target images in the class-domain of the base dataset with a fixed pretrained model might be sub-optimal. We propose a simple dynamic distillation-based approach to facilitate unlabeled images from the novel/base dataset. We impose consistency regularization by calculating predictions from the weakly-augmented versions of the unlabeled images from a teacher network and matching it with the strongly augmented versions of the same images from a student network. The parameters of the teacher network are updated as exponential moving average of the parameters of the student network. We show that the proposed network learns representation that can be easily adapted to the target domain even though it has not been trained with target-specific classes during the pretraining phase. Our model outperforms the current state-of-the art method by 4.4% for 1-shot and 3.6% for 5-shot classification in the BSCD-FSL benchmark, and also shows competitive performance on traditional in-domain few-shot learning task.
Open AccessDynamic Distillation Network for Cross-Domain Few-Shot Recognition with Unlabeled Data
Most existing works in few-shot learning rely on meta-learning the network on a large base dataset which is typically from the same domain as the target dataset. We tackle the problem of cross-domain few-shot learning where there is a large shift between the base and target domain. The problem of cross-domain few-shot recognition with unlabeled target data is largely unaddressed in the literature. STARTUP was the first method that tackles this problem using self-training. However, it uses a fixed teacher pretrained on a labeled base dataset to create soft labels for the unlabeled target samples. As the base dataset and unlabeled dataset are from different domains, projecting the target images in the class-domain of the base dataset with a fixed pretrained model might be sub-optimal. We propose a simple dynamic distillation-based approach to facilitate unlabeled images from the novel/base dataset. We impose consistency regularization by calculating predictions from the weakly-augmented versions of the unlabeled images from a teacher network and matching it with the strongly augmented versions of the same images from a student network. The parameters of the teacher network are updated as exponential moving average of the parameters of the student network. We show that the proposed network learns representation that can be easily adapted to the target domain even though it has not been trained with target-specific classes during the pretraining phase. Our model outperforms the current state-of-the art method by 4.4% for 1-shot and 3.6% for 5-shot classification in the BSCD-FSL benchmark, and also shows competitive performance on traditional in-domain few-shot learning task.
- 1 June 2015
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-4
IBM system G Social Media Solution: Analyze multimedia content, people, and network dynamics in context
We present IBM System G Social Media Solution, which includes a suite of applications designed for in-context monitoring, exploration, and analysis of social multimedia content as well as related people and network dynamics. Each individual application focuses on a unique aspect of social media data analysis in relevant context; collectively, they provide a comprehensive set of tools for exploring and analyzing real-time and historical social media data at large scale. The solution is empowered by a unified data management platform, based on a property graph model, to efficiently handle a large variety of social media applications.
- 1 June 2015
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-4
IBM system G Social Media Solution: Analyze multimedia content, people, and network dynamics in context
We present IBM System G Social Media Solution, which includes a suite of applications designed for in-context monitoring, exploration, and analysis of social multimedia content as well as related people and network dynamics. Each individual application focuses on a unique aspect of social media data analysis in relevant context; collectively, they provide a comprehensive set of tools for exploring and analyzing real-time and historical social media data at large scale. The solution is empowered by a unified data management platform, based on a property graph model, to efficiently handle a large variety of social media applications.
- Open Access
- 2 November 2013
- Published by Wiley in BioMed Research International
- Vol. 2013, 1-9
Analysis of Risk Factors for First Seizure after Stroke in Chinese Patients
The aim of this study is to assess related risk factors and predict early- and late-onset seizure after first-ever stroke. A total of 2474 consecutive patients with initial stroke in China from 1997 to 2007 were retrospectively investigated, in which, 24 clinical and radiological indexes were used for evaluation. Odds ratio (OR) and 95% confidence interval (CI) were calculated by logistic regression. A total of 232 (11.1%) of patients developed seizures during a mean follow-up period of 18 months, with 123 experiencing early-onset and 109 late-onset seizure. The independent risk factors for early-onset seizure were large lesion (), subarachnoid hemorrhage (), initial hyponatremia (), and cortical involvement (). The independent risk factors for late-onset seizure were cortical involvement () and large lesion (). These results demonstrated that the risk factors for early seizure after stroke are large lesion, subarachnoid hemorrhage, and cortical involvement. Surprisingly, hyponatremia also predicts seizure in stroke patients. Cortical involvement is a major risk factor for late-onset seizure after stroke.- 2 November 2013
- Published by Wiley in BioMed Research International
- Vol. 2013, 1-9
Open AccessAnalysis of Risk Factors for First Seizure after Stroke in Chinese Patients
The aim of this study is to assess related risk factors and predict early- and late-onset seizure after first-ever stroke. A total of 2474 consecutive patients with initial stroke in China from 1997 to 2007 were retrospectively investigated, in which, 24 clinical and radiological indexes were used for evaluation. Odds ratio (OR) and 95% confidence interval (CI) were calculated by logistic regression. A total of 232 (11.1%) of patients developed seizures during a mean follow-up period of 18 months, with 123 experiencing early-onset and 109 late-onset seizure. The independent risk factors for early-onset seizure were large lesion (), subarachnoid hemorrhage (), initial hyponatremia (), and cortical involvement (). The independent risk factors for late-onset seizure were cortical involvement () and large lesion (). These results demonstrated that the risk factors for early seizure after stroke are large lesion, subarachnoid hemorrhage, and cortical involvement. Surprisingly, hyponatremia also predicts seizure in stroke patients. Cortical involvement is a major risk factor for late-onset seizure after stroke.