A5042805005- Advanced Image Processing Techniques
- Advanced Vision and Imaging
- Image Processing Techniques and Applications
- Video Surveillance and Tracking Methods
- Image and Signal Denoising Methods
- Image Enhancement Techniques
- Visual Attention and Saliency Detection
- Domain Adaptation and Few-Shot Learning
- Adversarial Robustness in Machine Learning
- CCD and CMOS Imaging Sensors
- Computer Graphics and Visualization Techniques
- Innovative concrete reinforcement materials
- Cancer-related molecular mechanisms research
- Advanced Neural Network Applications
- Recycled Aggregate Concrete Performance
- Multimodal Machine Learning Applications
- Innovations in Concrete and Construction Materials
- Mathematical Biology Tumor Growth
- Geographic Information Systems Studies
- Immune Cell Function and Interaction
- Structural Behavior of Reinforced Concrete
- Advanced Data Compression Techniques
- Biochemical and biochemical processes
- Cellular Mechanics and Interactions
- COVID-19 diagnosis using AI
University of South China
2022-2025
Jiangsu University of Science and Technology
2024
Jining Medical University
2022-2024
Guangzhou Automobile Group (China)
2023
Guangdong University of Technology
2017-2022
Rice University
2021
Seoul National University
2021
Image deblurring is a classical computer vision problem that aims to recover sharp image from blurred image. To solve this problem, existing methods apply the Encode-Decode architecture design complex networks make good performance. However, most of these use repeated up-sampling and down-sampling structures expand receptive field, which results in texture information loss during sampling process some them multiple stages lead difficulties with convergence. Therefore, our model uses dilated...
Motion blur is a common photography artifact in dynamic environments that typically comes jointly with the other types of degradation. This paper reviews NTIRE 2021 Challenge on Image Deblurring. In this challenge report, we describe specifics and evaluation results from 2 competition tracks proposed solutions. While both aim to recover high-quality clean image blurry image, different artifacts are involved. track 1, images low resolution while compressed JPEG format. each competition, there...
This paper summarizes the 2nd NTIRE challenge on stereo image super-resolution (SR) with a focus new solutions and results. The task of is to super-resolve low-resolution pair high-resolution one magnification factor ×4. Compared single SR, major this lies in how exploit additional information another viewpoint maintain consistency has 3 tracks, including track distortion (e.g., PSNR) bicubic degradation, perceptual quality LPIPS) as well real degradations. In total, 175, 93, 103...
Abstract This study focused on the durability of basalt fibers, glass basalt‐fiber‐reinforced polymer (BFRP) bars, and glass‐fiber‐reinforced (GFRP) bars in an artificial seawater. The specimens were immersed seawater at 20, 40, 60°C; specimen distilled water a temperature 60°C was also used as reference. tensile strength monofilament fibers for 90 days decreases by 61% 59% respectively, showing most significant degradation; brittleness increases due to decomposition their sizing agent....
Low-level radioactive wastewater contains not only radionuclides but also a variety of inorganic and organic pollutants, which have an increasingly profound negative impact on the environment humans. Therefore, diverse technologies, including chemical precipitation, evaporation concentration, electrochemistry, other methods purifying been developed, these strategies generally suffer from secondary pollution or unsatisfactory purification effects. To further improve efficiency with minimized...
Recent papers have suggested that transfer learning can outperform sophisticated meta-learning methods for few-shot image classification. We take this hypothesis to its logical conclusion, and suggest the use of an ensemble high-quality, pre-trained feature extractors show experimentally a library combined with simple feed-forward network learned L2-regularizer be excellent option solving cross-domain Our experimental results approach far outperforms several well-established algorithms.
Quercetin, a flavonoid that is present in vegetables and fruits, has been found to have anti-inflammatory effects. However, the mechanism by which it inhibits colitis uncertain. This study aimed explore effect pharmacological of quercetin on dextran sodium sulfate (DSS)-induced ulcerative (UC). Mice were given 4% (w/v) DSS solution drink for 7 days, followed regular water following 5 days. Pharmacological mechanisms predicted network pharmacology. High-throughput 16S rDNA sequencing was...
In the real world, degradation of images taken under haze can be quite complex, where spatial distribution is varied from image to image. Recent methods adopt deep neural networks recover clean scenes hazy directly. However, due paradox caused by variation captured and fixed parameters current networks, generalization ability recent dehazing on real-world not ideal.To address problem modeling degradation, we propose solve this perceiving density for uneven distribution. We a novel Separable...
Stereo image super-resolution aims to improve the quality of high-resolution stereo pairs by exploiting complementary information across views. To attain superior performance, many methods have prioritized designing complex modules fuse similar views, yet overlooking importance intra-view for reconstruction. It also leads problems wrong texture in recovered images. address this issue, we explore interdependencies between various hierarchies from and propose a novel method, named Cross-View...
Video restoration networks aim to restore high-quality frame sequences from degraded ones. However, traditional video methods heavily rely on temporal modeling operators or optical flow estimation, which limits their versatility. The of this work is present a novel approach for that eliminates inefficient and pixel-level feature alignment in the network architecture. proposed method, Sequential Affinity Learning Network (SALN), designed based an affinity mechanism establishes direct...
In this paper, we propose a Convolutional Neural Network(CNN) model by combining Additive Margin Softmax with Center Loss for face recognition, which is denoted as AMS-CL. AMS-CL adopts the CNN network structure of ResNet50, pretrained on FaceScrub dataset. particular, exploits (i.e., AM-softmax) to increase inter-class distance, avoiding overconfidence classifications. Furthermore, regularization term center loss introduced decrease intra-class distance simultaneously. Therefore, keeps...
The promise of Deep Neural Network (DNN) powered Internet Thing (IoT) devices has motivated a tremendous demand for automated solutions to enable fast development and deployment efficient (1) DNNs equipped with instantaneous accuracy-efficiency trade-off capability accommodate the time-varying resources at IoT (2) dataflows optimize DNNs' execution efficiency on different devices. Therefore, we propose InstantNet automatically generate deploy instantaneously switchable-precision networks...
Dynamic texture describes the images sequence that continuously demonstrates movement of pixel's intensity change patterns in time. Two motion features, namely average radius and coherence index are extracted to find different characteristics fire smoke. Since, vectors rapid change, whilst has been developed assess coherent change. We propose a cutting-edge periodic series analysis characterize smoke dynamic textures. Our experimental results demonstrate pattern Besides, periodicity...
The promise of Deep Neural Network (DNN) powered Internet Thing (IoT) devices has motivated a tremendous demand for automated solutions to enable fast development and deployment efficient (1) DNNs equipped with instantaneous accuracy-efficiency trade-off capability accommodate the time-varying resources at IoT (2) dataflows optimize DNNs' execution efficiency on different devices. Therefore, we propose InstantNet automatically generate deploy instantaneously switchable-precision networks...
Stereo image super-resolution aims to improve the quality of high-resolution stereo pairs by exploiting complementary information across views. To attain superior performance, many methods have prioritized designing complex modules fuse similar views, yet overlooking importance intra-view for reconstruction. It also leads problems wrong texture in recovered images. address this issue, we explore interdependencies between various hierarchies from and propose a novel method, named...
The development of online high-definition maps is significant since they provide real-time, accurate, and updatable geographic information for location-based applications, such as autonomous driving intelligent transportation, thus improving the performance reliability these applications. Previous works, VectorMapNet MapTR, show that direct model generation vectorized HD a promising solution. However, methods did not take into account usage global semantic to improve map construction...
The stability analysis of a chemotaxis model with bistable growth term in both unbounded and bounded domains is studied analytically. By the global bifurcation theorem, we identify full parameter regimes which steady state occurs.