A5089406697- Medical Imaging and Analysis
- Scoliosis diagnosis and treatment
- Medical Image Segmentation Techniques
- Ultrasound Imaging and Elastography
- Building Energy and Comfort Optimization
- Emotion and Mood Recognition
- AI in cancer detection
- Traffic Prediction and Management Techniques
- Innovative Microfluidic and Catalytic Techniques Innovation
- Image Retrieval and Classification Techniques
- Microfluidic and Capillary Electrophoresis Applications
- Spinal Fractures and Fixation Techniques
- BIM and Construction Integration
- Electrowetting and Microfluidic Technologies
- Brain Tumor Detection and Classification
- Transportation Planning and Optimization
- Image Processing Techniques and Applications
- Advanced Technologies in Various Fields
- Advanced Image Fusion Techniques
- Gait Recognition and Analysis
- Ammonia Synthesis and Nitrogen Reduction
- 3D Surveying and Cultural Heritage
- Face and Expression Recognition
- Face recognition and analysis
- Dental Radiography and Imaging
Hong Kong Polytechnic University
2020-2024
Hanshan Normal University
2024
University of California, Berkeley
2023
University of Technology Sydney
2021
Facial expression recognition (FER) is of great interest to the current studies human-computer interaction. In this paper, we propose a novel geometry-guided facial framework, based on graph convolutional networks and transformers, perform effective emotion from videos. Specifically, detect utilize landmarks construct spatial-temporal graph, both landmark coordinates local appearance, for representing sequence. The blocks transformer modules are employed produce high-semantic emotion-related...
Volume Projection Imaging from ultrasound data is a promising technique to visualize spine features and diagnose Adolescent Idiopathic Scoliosis. In this paper, we present novel multi-task framework reduce the scan noise in volume projection images segment different simultaneously, which provides an appealing alternative for intelligent scoliosis assessment clinical applications. Our proposed consists of two streams: i) A removal stream based on generative adversarial networks, aims achieve...
Learning discriminative representations with good robustness from facial observations serves as a fundamental step towards intelligent expression recognition (FER). In this article, we propose novel geometry-aware FER framework to boost the performance based on both geometric and appearance knowledge. Specifically, an encoding strategy for landmarks, adopt graph convolutional network (GCN) fully explore structural information of components behind different expressions. A neural (CNN) is...
Scoliosis is a widespread medical condition where the spine becomes severely deformed and bends over time. It mostly affects young adults may have permanent impact on them. A periodic assessment, using suitable modality, necessary for its early detection. Conventionally, usually employed modalities include X-ray MRI, which employ ionising radiation are expensive. Hence, non-radiating 3D ultrasound imaging technique has been developed as safe economic alternative. However, produces...
In medical image analysis, anatomical landmarks usually contain strong prior knowledge of their structural information. this paper, we propose to promote landmark localization by modeling the underlying distribution via normalizing flows. Specifically, introduce flow-based as a learnable objective function into regression-based framework. Moreover, employ an integral operation make mapping from heatmaps coordinates differentiable further enhance heatmap-based with learned prior. Our proposed...
Microfluidic-based ultramicroscale chemistry experiments involve the manipulation and consumption of reagents on a microliter or submicroliter scale in microfluidic device. This article presents pedagogical functions systems, including green education, stimulation students' interest learning about chemistry, interdisciplinary scientific knowledge. We also review development educational courses platforms for implementing education examine teaching forms demonstrations, laboratory experiments,...
3D ultrasound imaging shows great promise for scoliosis diagnosis thanks to its low-costing, radiation-free and real-time characteristics. The key accessing by is accurately segment the bone area measure degree based on symmetry of features. images tend contain many speckles regular occlusion noise which difficult, tedious time-consuming experts find out bony feature. In this paper, we propose a robust feature segmentation method U-net structure spine Volume Projection Imaging (VPI) images....
Ultrasound volume projection imaging has shown great promise to visualize spine features and diagnose scoliosis thanks its harmlessness, cheapness, efficiency. The key measuring deformity assessing is accurately segment the bone features. In this paper, we propose a novel structure-affinity dual attention-based network (SADANet) for effective segmentation. Global channel attention module spatial criss-cross are combined in parallel manner generate rich global context of images. Meanwhile,...
Building archetypes, representative models of building stock, are crucial for precise energy simulations in Urban Energy Modeling. The current widely adopted archetypes developed on a nationwide scale, potentially neglecting the impact local buildings' geometric specificities. We present Multi-scale Archetype Representation Learning (MARL), an approach that leverages representation learning to extract features from specific stock. Built upon VQ-AE, MARL encodes footprints and purifies...
Automatic spine segmentation, based on ultrasound volume projection imaging (VPI), is of great value in clinical applications to diagnose scoliosis teenagers. In this paper, we propose a novel framework improve the segmentation accuracy images via structure-enhanced attentive learning. Since bones contain strong prior knowledge their shapes and positions VPI images, encode information into semantic representations an manner. We first revisit self-attention mechanism representation learning,...
Ultrasound volume projection imaging (VPI) has shown to be appealing from a clinical perspective, because of its harmlessness, flexibility, and efficiency in scoliosis assessment. However, the limitations hardware devices degrade resultant image content with strong structured noise. Owing unavailability reference data unpredictable degradation model, VPI recovery is challenging problem. In this paper, we propose novel framework learn noise removal unpaired samples. We introduce attention...
As the global population and urbanization expand, building sector has emerged as predominant energy consumer carbon emission contributor. The need for innovative Urban Building Energy Modeling grows, yet existing archetypes often fail to capture unique attributes of local buildings nuanced distinctions between different cities, jeopardizing precision modeling. This paper presents an alternative tool employing self-supervised learning distill complex geometric data into representative,...
Ultrasound curve angle (UCA) measurement provides a radiation-free and reliable evaluation for scoliosis based on ultrasound imaging. However, degraded image quality, especially in difficult-to-image patients, can prevent clinical experts from making confident measurements, even leading to misdiagnosis. In this paper, we propose multi-stage enhancement framework that models high-quality distribution via diffusion-based model. Specifically, integrate the underlying morphological information...
Heterogeneous catalysts possess many advantages over single-phase catalysts. However, how to tune the heterogeneous interface state and density in heterostructures optimize their electrocatalytic performance is still a challenge. Here, we propose method for preparation of NiS-NiS2 heterostructure by vulcanization nickel foam at high temperature，which can effectively weight ratio NiS NiS2 corresponding urea oxidation reaction (UOR), hydrogen evolution (HER) overall splitting performance. The...
The potential of digital-twin technology, involving the creation precise digital replicas physical objects, to reshape AR experiences in 3D object tracking and localization scenarios is significant. However, enabling robust dynamic mobile environments remains a formidable challenge. These often require more pose estimator capable handling inherent sensor-level measurement noise. In this paper, recognizing challenges comprehensive solutions existing literature, we propose transformer-based...
Building archetypes, representative models of building stock, are crucial for precise energy simulations in Urban Energy Modeling. The current widely adopted archetypes developed on a nationwide scale, potentially neglecting the impact local buildings' geometric specificities. We present Multi-scale Archetype Representation Learning (MARL), an approach that leverages representation learning to extract features from specific stock. Built upon VQ-AE, MARL encodes footprints and purifies...
In this paper, we propose a machine learning-based approach to address the lack of ability for designers optimize urban land use planning from perspective vehicle travel demand. Research shows that our computational model can help quickly obtain feedback on demand, which includes its total amount and temporal distribution based function designed by designers. It also assists in design optimization evaluation travel. We city information hours traveled (VHT) collecting point-of-interest (POI)...
This paper includes a review of current state the art 6d pose estimation methods, as well discussion which method should be used in two types architectural design scenarios. Taking latest research Gen6d an example, we make qualitative assessment openset methods terms application level, prediction speed, resistance to occlusion, accuracy, environmental interference, etc. In addition, try combine 6D and building wind environment create tangible approach, discuss limitations point out direction...