A5000467703- Medical Image Segmentation Techniques
- AI in cancer detection
- Advanced Vision and Imaging
- Generative Adversarial Networks and Image Synthesis
- Advanced Neural Network Applications
- Advanced Image and Video Retrieval Techniques
- Image Retrieval and Classification Techniques
- Cell Image Analysis Techniques
- Medical Imaging Techniques and Applications
- Robotics and Sensor-Based Localization
- 3D Shape Modeling and Analysis
- Radiomics and Machine Learning in Medical Imaging
- Cervical Cancer and HPV Research
- Human Pose and Action Recognition
- Image and Object Detection Techniques
- Optical measurement and interference techniques
- Advanced Image Processing Techniques
- Computer Graphics and Visualization Techniques
- Cellular Mechanics and Interactions
- Visual Attention and Saliency Detection
- Digital Imaging for Blood Diseases
- Multimodal Machine Learning Applications
- Domain Adaptation and Few-Shot Learning
- Medical Imaging and Analysis
- Spectroscopy Techniques in Biomedical and Chemical Research
Gannan Medical University
2025
Pennsylvania State University
2019-2024
Royal Society of Chemistry
2024
Centre National pour la Recherche Scientifique et Technique (CNRST)
2024
Penn State Milton S. Hershey Medical Center
2023
University of Chicago
2023
Soochow University
2023
Southern Medical University
2022
Shenzhen Maternity and Child Healthcare Hospital
2022
Weifang Medical University
2020-2021
Reliable estimation of visual saliency allows appropriate processing images without prior knowledge their contents, and thus remains an important step in many computer vision tasks including image segmentation, object recognition, adaptive compression. We propose a regional contrast based extraction algorithm, which simultaneously evaluates global differences spatial coherence. The proposed algorithm is simple, efficient, yields full resolution maps. Our consistently outperformed existing...
Automatic estimation of salient object regions across images, without any prior assumption or knowledge the contents corresponding scenes, enhances many computer vision and graphics applications. We introduce a regional contrast based detection algorithm, which simultaneously evaluates global differences spatial weighted coherence scores. The proposed algorithm is simple, efficient, naturally multi-scale, produces full-resolution, high-quality saliency maps. These maps are further used to...
In this paper, we propose an Attentional Generative Adversarial Network (AttnGAN) that allows attention-driven, multi-stage refinement for fine-grained text-to-image generation. With a novel attentional generative network, the AttnGAN can synthesize details at different sub-regions of image by paying attentions to relevant words in natural language description. addition, deep multimodal similarity model is proposed compute image-text matching loss training generator. The significantly...
Although Generative Adversarial Networks (GANs) have shown remarkable success in various tasks, they still face challenges generating high quality images. In this paper, we propose Stacked (StackGANs) aimed at high-resolution photo-realistic First, a two-stage generative adversarial network architecture, StackGAN-v1, for text-to-image synthesis. The Stage-I GAN sketches the primitive shape and colors of scene based on given text description, yielding low-resolution Stage-II takes results...
Although promising results have been achieved in the areas of traffic-sign detection and classification, few works provided simultaneous solutions to these two tasks for realistic real world images. We make contributions this problem. Firstly, we created a large benchmark from 100000 Tencent Street View panoramas, going beyond previous benchmarks. It provides images containing 30000 instances. These cover variations illuminance weather conditions. Each is annotated with class label, its...
Most convolutional neural networks (CNNs) lack midlevel layers that model semantic parts of objects. This limits CNN-based methods from reaching their full potential in detecting and utilizing small recognition. Introducing such mid-level can facilitate the extraction part-specific features which be utilized for better recognition performance. is particularly important domain fine-grained In this paper, we propose a new CNN architecture integrates part detection abstraction (SPDACNN)...
Abstract Understanding the breakdown mechanisms of polymer-based dielectrics is critical to achieving high-density energy storage. Here a comprehensive phase-field model developed investigate electric, thermal, and mechanical effects in process dielectrics. High-throughput simulations are performed for P(VDF-HFP)-based nanocomposites filled with nanoparticles different properties. Machine learning conducted on database from high-throughput produce an analytical expression strength, which...
Synthesizing high-quality images from text descriptions is a challenging problem in computer vision and has many practical applications. Samples generated by existing text-to-image approaches can roughly reflect the meaning of given descriptions, but they fail to contain necessary details vivid object parts. In this paper, we propose Stacked Generative Adversarial Networks (StackGAN) generate 256x256 photo-realistic conditioned on descriptions. We decompose hard into more manageable...
Conditional image-to-video (cI2V) generation aims to synthesize a new plausible video starting from an image (e.g., person's face) and condition action class label like smile). The key challenge of the cI2V task lies in simultaneous realistic spatial appearance temporal dynamics corresponding given condition. In this paper, we propose approach for using novel latent flow diffusion models (LFDM) that optical sequence space based on warp image. Compared previous direct-synthesis-based works,...
We present a novel, variational and statistical approach for shape registration. Shapes of interest are implicitly embedded in higher-dimensional space distance transforms. In this implicit embedding space, registration is formulated hierarchical manner: the mutual information criterion supports various transformation models optimized to perform global registration; then, B-spline-based incremental free form deformations (IFFD) model used minimize sum-of-squared-differences (SSD) measure...
We use open active contours to quantify cytoskeletal structures imaged by fluorescence microscopy in two and three dimensions. developed an interactive software tool for segmentation, tracking, visualization of individual fibers. Open are parametric curves that deform minimize the sum external energy derived from image internal bending stretching energy. The generates (i) forces attract contour toward central bright line a filament image, (ii) stretch ends ridges. Images simulated...
Repeated elements are ubiquitous and abundant in both manmade natural scenes. Editing such images while preserving the repetitions their relations is nontrivial due to overlap, missing parts, deformation across instances, illumination variation, etc. Manually enforcing laborious error-prone. We propose a novel framework where user scribbles used guide detection extraction of repeated elements. Our process, which based on boundary band method, robustly extracts along with deformations. The...
This paper investigates a new learning formulation called dynamic group sparsity. It is natural extension of the standard sparsity concept in compressive sensing, and motivated by observation that some practical sparse data nonzero coefficients are often not random but tend to be clustered. Intuitively, better results can achieved these cases reasonably utilizing both clustering priors. Motivated this idea, we have developed greedy recovery algorithm, which prunes residues iterative process...
In this work, we propose to resolve the issue existing in current deep learning based organ segmentation systems that they often produce results do not capture overall shape of target and lack smoothness. Since there is a rigorous mapping between Signed Distance Map (SDM) calculated from object boundary contours binary map, exploit feasibility SDM directly medical scans. By converting task into predicting an SDM, show our proposed method retains superior performance has better smoothness...
Abstract Filamentous biopolymer networks in cells and tissues are routinely imaged by confocal microscopy. Image analysis methods enable quantitative study of the properties these curvilinear networks. However, software tools to quantify geometry topology often dense 3D localize network junctions scarce. To fill this gap, we developed a new tool called “SOAX”, which can accurately extract centerlines identify using Stretching Open Active Contours (SOACs). It provides an open-source,...
Although Generative Adversarial Networks (GANs) have shown remarkable success in various tasks, they still face challenges generating high quality images. In this paper, we propose Stacked (StackGAN) aiming at high-resolution photo-realistic First, a two-stage generative adversarial network architecture, StackGAN-v1, for text-to-image synthesis. The Stage-I GAN sketches the primitive shape and colors of object based on given text description, yielding low-resolution Stage-II takes results...
The study of Alzheimer's disease (AD), the most common cause dementia, faces challenges in terms understanding cause, monitoring pathogenesis, and developing early diagnoses effective treatments. Rapid accurate identification AD biomarkers brain is critical to providing key insights into facilitating development diagnosis methods. In this work, we developed a platform that enables rapid screening by employing graphene-assisted Raman spectroscopy machine learning interpretation transgenic...
Abstract Synthesis and re‐targeting of facial expressions is central to animation often involves significant manual work in order achieve realistic expressions, due the difficulty capturing high quality dynamic expression data. In this paper we address fundamental issues regarding use dense 3‐D data samples undergoing motions at video speeds, e.g. human expressions. utilize such for motion analysis re‐targeting, correspondences must be established between different frames same faces as well...