A5100645300- Advanced Neural Network Applications
- Reservoir Engineering and Simulation Methods
- Domain Adaptation and Few-Shot Learning
- Multimodal Machine Learning Applications
- Hydraulic Fracturing and Reservoir Analysis
- 3D Shape Modeling and Analysis
- 3D Surveying and Cultural Heritage
- Advanced Image and Video Retrieval Techniques
- Drilling and Well Engineering
- Visual Attention and Saliency Detection
- Software Reliability and Analysis Research
- Hydrocarbon exploration and reservoir analysis
- Reliability and Maintenance Optimization
- Oil and Gas Production Techniques
- Computer Graphics and Visualization Techniques
- Robotic Mechanisms and Dynamics
- Risk and Safety Analysis
- Occupational Health and Safety Research
- Remote Sensing and LiDAR Applications
- Adversarial Robustness in Machine Learning
- Advancements in Battery Materials
- Enhanced Oil Recovery Techniques
- Soft Robotics and Applications
- Neural Networks and Applications
- Structural Health Monitoring Techniques
Beijing University of Chemical Technology
2012-2025
Nanjing University of Chinese Medicine
2025
UCSI University
2025
Technical University of Denmark
2025
Southeast University
2022-2024
Sinopec (China)
2024
Shanghai University of Engineering Science
2021-2024
University of Amsterdam
2023-2024
Amsterdam University of the Arts
2024
Toyota Technological Institute at Chicago
2024
The crux of semi-supervised semantic segmentation is to assign adequate pseudo-labels the pixels unlabeled images. A common practice select highly confident predictions as pseudo ground-truth, but it leads a problem that most may be left unused due their unreliability. We argue every pixel matters model training, even its prediction ambiguous. Intuitively, an unreliable get confused among top classes (i.e., those with highest probabilities), however, should about not belonging remaining...
Summary History matching is a typical inverse problem that adjusts the uncertainty parameters of reservoir numerical model with limited dynamic response data. In most situations, various parameter combinations can result in same data fit, termed as nonuniqueness inversion. It desirable to find many global or local optima possible single optimization run, which may help reveal distribution posterior space, particularly important for robust optimization, risk analysis, and decision making...
In clinical acupuncture practice, needle twirling (NT) and retention (NR) are strategically combined to achieve different therapeutic effects, highlighting the importance of distinguishing between states. Scalp EEG has been proven significantly relevant brain activity stimulation. this work, we designed an paradigm collect scalp study differences in changes during Since deep learning (DL) increasingly used analysis, propose Acupuncture Transformer Detector (ATD), a model based on...
Video Instance Segmentation (VIS) aims at segmenting and categorizing objects in videos from a closed set of training categories, lacking the generalization ability to handle novel categories real-world videos. To address this limitation, we make following three contributions. First, introduce task Open-Vocabulary Segmentation, which simultaneously segment, track, classify open-set including unseen during training. Second, benchmark VIS, collect Large-Vocabulary dataset (LV-VIS), that...
Self-training has shown great potential in semi-supervised learning. Its core idea is to use the model learned on labeled data generate pseudo-labels for unlabeled samples, and turn teach itself. To obtain valid supervision, active attempts typically employ a momentum teacher pseudo-label prediction yet observe confirmation bias issue, where incorrect predictions may provide wrong supervision signals get accumulated training process. The primary cause of such drawback that prevailing...
With the accelerated depletion of non-renewable resources and increased demand for lithium batteries, green recycling has become a key issue nowadays. In this study, effects mass ratio potassium persulfate to active material battery cathode material, roasting temperature, time, liquid-solid leaching time on rate lithium, cobalt, nickel manganese were investigated. For lithium-cobalt oxide materials, optimal conditions K₂S₂O₇ LiCoO₂ 2:3, temperature 700 °C 60 min, 98.51% selective 99.86%....
In this paper, we propose variational prototype inference to address few-shot semantic segmentation in a probabilistic framework. A latent variable model infers the distribution of that is treated as variable. We formulate optimization problem, which established with an amortized network based on auto-encoder architecture. The modeling enhances its generalization ability handle inherent uncertainty caused by limited data and huge intra-class variations objects. Moreover, it offers principled...
Scour is one of the primary reasons for collapse bridges, as it severely reduces stability their piers. This study developed a model updating method based on an improved genetic algorithm (GA) to identify scour depth, combined with parallel computing technique accelerate procedure. The proposed GA adopted gradient-like calculation mutation operation, while population in was initialized exponential distribution. A finite element continuous girder bridge, whose soil-structure interaction...
Non-maximum suppression (NMS) is widely used in object detection pipelines for removing duplicated bounding boxes. The inconsistency between the confidence NMS and real localization seriously affects performance. Prior works propose to predict Intersection-over-Union (IoU) boxes corresponding ground-truths improve NMS, while accurately predicting IoU still a challenging problem. We argue that complex definition of feature misalignment make it difficult accurately. In this paper, we novel...
Petroleum is widely used as one of the important energy sources. Engineers need to continuously adjust strategies maximize oil recovery during development. Then strategy adjustment requires a lot cost, so it necessary select optimal through production forecasting. In this paper, self-defined Double-channel Heterogeneous Dynamical Graph network (DHDG) proposed deal with problem. It can receive different development predict future indices base on that. Particularly, concept graph for first...
Summary Surrogate-assisted evolutionary algorithms (SAEAs) have become a popular approach for solving reservoir production optimization problems. The radial-basis-function network (RBFN) is robust surrogate model technology suitable development with numerous wells and long lifetime. There are several types of basis functions constructing RBFN models. However, existing research shows that selecting the function competitive performance current problem challenging without prior knowledge. In...
Abstract The tension distribution problem of cable-driven parallel robots is inevitable in real-time control. Currently, iterative algorithms or geometric are commonly used to solve this problem. Iterative difficult improve performance, and the obtained by may not be continuous. In paper, a novel method for four-cable, 3-DOF proposed based on wave equation. calculated continuous differentiable, without need computation centroid calculations, thus exhibiting good performance. Furthermore,...
Summary Reservoir connectivity analysis plays an essential role in controlling water cut the middle and later stages of reservoir development. The traditional methods, such as well test tracer, may result interruption high development costs. Analyzing through history data is advisable alternative method because fluctuation reflects interwell interference. However, most former data-driven capacitance resistance model (CRM), estimate using formulas relatively simple forms, leading to...
Summary Evolutionary transfer optimization (ETO) algorithms with the ability to learn from past tasks have made breakthroughs in more and fields. When experience embedded is properly utilized, search performance will be greatly improved compared starting scratch. Autoencoding evolutionary (AEES) an efficient ETO paradigm proposed recent years. The solutions of each task are configured as input output a single-layer denoising autoencoder (DAE), across-problem mapping established by minimizing...
Automatic check-out (ACO) emerges as an integral component in recent self-service retailing stores, which aims at automatically detecting and counting the randomly placed products upon a platform. Existing data-driven works still have difficulties generalizing to real-world retail product scenarios, since (1) real images are hard collect or cover all their possible layouts, (2) rapid updating of list leads frequent tedious re-training models. To overcome these obstacles, we contribute...
As one of the most fundamental and challenging problems in computer vision, object detection has received tremendous attention recent years. Its development over past two decades can be seen as a microcosm history vision. The goal face is to find corresponding positions all faces image. output algorithm coordinates circumscribing rectangle image, may also include information such attitude tilt angle. Algorithms are constantly updated address higher accuracy, lower computational complexity,...