A5100367807- Software Engineering Research
- Music and Audio Processing
- Digital Media Forensic Detection
- Speech and Audio Processing
- Software System Performance and Reliability
- Data Visualization and Analytics
- Speech Recognition and Synthesis
- Human Mobility and Location-Based Analysis
- Remote-Sensing Image Classification
- Advanced Image Fusion Techniques
- Microplastics and Plastic Pollution
- Advanced Neural Network Applications
- Scientific Computing and Data Management
- Video Analysis and Summarization
- Multimedia Communication and Technology
- Human Pose and Action Recognition
- Medical Image Segmentation Techniques
- Topic Modeling
- Advanced Image and Video Retrieval Techniques
- Image and Signal Denoising Methods
- Land Use and Ecosystem Services
- Remote Sensing in Agriculture
- Logic, programming, and type systems
- Product Development and Customization
- Statistics Education and Methodologies
Wuhan University
2022-2024
Peking University
2014-2024
Shenzhen University
2024
Ministry of Natural Resources
2024
Peking University Cancer Hospital
2024
CHN Energy (China)
2024
East China Normal University
2023
Peking University Shenzhen Hospital
2023
Northeastern University
2022
Microsoft Research (United Kingdom)
2022
Abstract Parkinson's disease (PD) diagnosis based on magnetic resonance imaging (MRI) is still challenging clinically. Quantitative susceptibility maps (QSM) can potentially provide underlying pathophysiological information by detecting the iron distribution in deep gray matter (DGM) nuclei. We hypothesized that learning (DL) could be used to automatically segment all DGM nuclei and use relevant features for a better differentiation between PD healthy controls (HC). In this study, we...
While visualizations play a crucial role in gaining insights from data, generating useful complex dataset is far an easy task. In particular, besides understanding the functionality provided by existing visualization libraries, desired also requires reshaping and aggregating underlying data as well composing different visual elements to achieve intended narrative. This paper aims simplify tasks automatically synthesizing required program simple sketches user. Specifically, given input set...
Information Theory of Cartography plays an important role in guiding map design, generalization, and evaluation, measurement information is the most basic topic this theory. However, there are many problems current methods, a long way to go form theoretically rigorous algorithm that can effectively depict spatial comprehensively consider feeling readers. Luckily, we now propose evaluation metric exhibits certain correlation with based on deep learning benefit actual cartography, term it as...
Abstract Damming profoundly affects downstream flow‐sediment regimes, altering channel bar dynamics and thereby affecting floods riverine biodiversity. Here, we investigate the response of bars to upstream damming by examining patterns, mechanisms, impacts in Middle Yangtze River (∼1,000 km). Over a decade post‐damming observational data reveal substantial increases revegetation Leaf Area Index. Shorter flood duration stable size collectively drive greening. Consequently, denser vegetation...
Oblique stripe noise widely appears in remote sensing images after image correction, exhibiting arbitrary tilt angles and parallel distribution. Due to its randomness lengths, oblique increases the difficulty of detection compared vertical or horizontal noise. For first time, we propose a group synthesis algorithms combining imaging mechanisms deep learning. To get controllable synthetic data for training model, two sample augmentation methods are presented by correction's with new linear...
This study applies variogram analyses of normalized difference vegetation index (NDVI) images derived from SPOT HRV obtained before and after the ChiChi earthquake in Chenyulan watershed, Taiwan, as well four large typhoons, to delineate spatial patterns, structures variability landscapes caused by these disturbances. The conditional Latin hypercube sampling approach was applied select samples multiple NDVI images. Kriging sequential Gaussian simulation with sufficient were then used...
Reward is deemed a performance reinforcer. The current study investigated how social and monetary reward anticipation affected cognitive control in 39 children, 40 adolescents, adults. We found that improved with age Simon task, the reaction time (RT) was modulated by magnitude. conflict monitoring process (target N2 amplitudes) of adolescents attentional processes P3 adults could be adjusted magnitude, suggesting were more sensitive to rewards compared children. magnitudes influenced neural...
Large language models (LLMs) have demonstrated an impressive ability to generate code for various programming tasks. In many instances, LLMs can a correct program task when given numerous trials. Consequently, recent trend is do large scale sampling of programs using model and then filtering/ranking the based on execution small number known unit tests select one candidate solution. However, these approaches assume that are safely execute generated (which arbitrary dangerous operations such...
Nanocellulose membranes could efficiently separate oily wastewater because of their super-hydrophilic and underwater super-oleophobic property nano-porous structure. However, the practical application storage nanocellulose is limited by low water permeation flux easy corrosion bacteria, respectively. Herein, with high permeate antibacterial were fabricated grafting tetrakis(hydroxymethyl) phosphonium chloride (THPC) onto surface TEMPO-oxidized tunicate cellulose nanofibers (TCNFs) via...
Java program adaptation between different APIs is a common task in software development. When an old API upgraded to incompatible new version, or when we want migrate application from one platform another platform, need adapt programs APIs. Although transformation tools have been developed automate the task, no tool ensures type safety transforming programs: given and any well-typed program, transformed result still well-typed. As matter of fact, it often observed that dedicated turns...
Code generation models can benefit data scientists' productivity by automatically generating code from context and text descriptions. An important measure of the modeling progress is whether a model generate that correctly execute to solve task. However, due lack an evaluation dataset directly supports execution-based evaluation, existing work relies on surface form similarity metrics (e.g., BLEU, CodeBLEU) for selection, which be inaccurate. To remedy this, we introduce ExeDS, execution...
While visualizations play a crucial role in gaining insights from data, generating useful complex dataset is far an easy task. Besides understanding the functionality provided by existing visualization libraries, desired also requires reshaping and aggregating underlying data as well composing different visual elements to achieve intended narrative. This paper aims simplify tasks automatically synthesizing required program simple sketches user. Specifically, given input set sketch that...
Naturalistic driving action recognition plays an important role in understanding drivers' distracted behaviors the traffic environment. The main challenge of this task is accurate localization temporal boundary for each behavior video. Although many methods can identify categories, it difficult to predict boundaries since actions same category usually present large intra-class variation. In paper, we introduce a Coarse-to-Fine Boundary Localization method called CFBL, which obtains...
Users often rely on GUIs to edit and interact with visualizations - a daunting task due the large space of editing options. As result, users are either overwhelmed by complex UI or constrained custom tailored, fixed subset options limited flexibility. Natural Language Interfaces (NLIs) emerging as feasible alternative for specify edits. However, NLIs forgo advantages traditional GUI: ability explore repeat edits see instant visual feedback. We introduce DynaVis, which blends natural language...
We assessed the performance of diffusion models for assessing response to neoadjuvant chemotherapy (NACT) using machine learning. Firstly, features were extracted from region interest on different parametric maps esophageal squamous cell carcinoma (ESCC) patients and changes parameters (Δ parameter) before after NACT (pre-NACT post-NACT) calculated. Then Δ-NACT pre-NACT built models. The results demonstrated that may be used predict efficacy in ESCC patients.
In this paper, a novel SAR image despeckling method based on the nonlocal centralized sparse representation (NCSR) is presented. NCSR model initially proposed for restoration, exhibiting excellent denoising performance nature images corrupted by additive Gaussian noise. However, images, whose speckle noise multiplicative and follows non-Gaussian distribution, are very different from images. Considering properties of speckle, pre-processed, then generalized likelihood ratio (GLR) criteria,...
Understanding multiple types of data source and correlating them make intelligence analysis a challenging task. Visual analytics techniques help analysts better comprehend the find correlations. We proposed visual system mainly designed for collecting evidence in analysis. Based on our system, we provided analytical solution IEEE VAST Challenge 2014 Grand Challenge. With collect information from spatial, temporal network data. Multiple methods are to generate wide range hypotheses test their...
Accurately predicting human mobility is crucial for understanding dynamics, addressing urban challenges, and improving planning. However, it still faces challenges due to the dynamic complex nature of mobility. Traditional methods rely on manual feature design statistical analysis, limiting accuracy generalizability predictions. In recent years, development deep learning techniques has provided new solutions. this study, we propose a Multi-perspective Spatiotemporal Context-aware Neural...
Fake audio detection is a growing concern and some relevant datasets have been designed for research. However, there no standard public Chinese dataset under complex conditions.In this paper, we aim to fill in the gap design fake (CFAD) studying more generalized methods. Twelve mainstream speech-generation techniques are used generate audio. To simulate real-life scenarios, three noise selected adding at five different signal-to-noise ratios, six codecs considered transcoding (format...