A5058329224- Hydrocarbon exploration and reservoir analysis
- Advanced Neural Network Applications
- Coal Properties and Utilization
- Advanced Vision and Imaging
- Hydraulic Fracturing and Reservoir Analysis
- Image Processing Techniques and Applications
- Atmospheric and Environmental Gas Dynamics
- 3D Surveying and Cultural Heritage
- Multimodal Machine Learning Applications
- Sulfur-Based Synthesis Techniques
- Human Pose and Action Recognition
- Speech and dialogue systems
- Radical Photochemical Reactions
- Optical measurement and interference techniques
- Robotics and Sensor-Based Localization
- Geoscience and Mining Technology
- Image Enhancement Techniques
- Catalytic C–H Functionalization Methods
- Computer Graphics and Visualization Techniques
- Domain Adaptation and Few-Shot Learning
- 3D Shape Modeling and Analysis
- Autonomous Vehicle Technology and Safety
- Food Supply Chain Traceability
- Advanced Photocatalysis Techniques
- Electrowetting and Microfluidic Technologies
University of Macau
2022-2025
Dalian University of Technology
2022-2025
City University of Macau
2024-2025
Sinopec (China)
2024
Dalian University
2022-2023
China University of Petroleum, East China
2018-2021
Dalian Maritime University
2021
University of Shanghai for Science and Technology
2018
Ansteel (China)
2018
China University of Petroleum, Beijing
2014
Abstract The full pore size distribution represents the integrated characteristics of micro‐nano pore‐throat systems in tight reservoirs. And it involves experiments different scales to fully analyze microscope properties. In this paper, we established a new approach for characterization through conducting high‐pressure mercury intrusion (HPMI) and low‐temperature nitrogen gas adsorption (LTN 2 GA) experiments. Meanwhile, studied petrology feature sandstones X‐ray diffraction (X‐rD) TESCAN...
Existing referring understanding tasks tend to involve the detection of a single text-referred object. In this paper, we propose new and general task, termed multi-object tracking (RMOT). Its core idea is employ language expression as semantic cue guide prediction tracking. To best our knowledge, it first work achieve an arbitrary number referent object predictions in videos. push forward RMOT, construct one benchmark with scalable expressions based on KITTI, named Refer-KITTI. Specifically,...
A new trend in the computer vision community is to capture objects of interest following flexible human command represented by a natural language prompt. However, progress using prompts driving scenarios stuck bottleneck due scarcity paired prompt-instance data. To address this challenge, we propose first object-centric prompt set for scenes within 3D, multi-view, and multi-frame space, named NuPrompt. It expands Nuscenes dataset constructing total 35,367 descriptions, each referring an...
PPh3 is shown to be an effective reagent for the deoxygenative reduction of allyl sulfones sulfides under photocatalytic conditions. Cross‐over experiment suggests that reaction proceeds in a sequential approach involving radical fragmentation and deoxygenation sulfonyl thiyl radical, which recombines with terminal alkene. This mild protocol exhibits broad functional group tolerance, its applications are explored through scale‐up reactions.
Cephalometric landmark detection is essential for orthodontic diagnostics and treatment planning. Nevertheless, the scarcity of samples in data collection extensive effort required manual annotation have significantly impeded availability diverse datasets. This limitation has restricted effectiveness deep learning-based methods, particularly those based on large-scale vision models. To address these challenges, we developed an innovative generation method capable producing cephalometric...
There are two crucial aspects of reliable autonomous driving systems: the reasoning behind decision-making and precision environmental perception. This paper introduces DME-Driver, a new system that enhances performance robustness by fully leveraging aspects. comprises main models. The first, Decision Maker, is responsible for providing logical instructions. second, Executor, receives these instructions generates precise control signals vehicles. To ensure explainable decisions, we build...
A new trend in the computer vision community is to capture objects of interest following flexible human command represented by a natural language prompt. However, progress using prompts driving scenarios stuck bottleneck due scarcity paired prompt-instance data. To address this challenge, we propose first object-centric prompt set for scenes within 3D, multi-view, and multi-frame space, named NuPrompt. It expands nuScenes dataset constructing total 40,147 descriptions, each referring an...
Adsorption is one of the most important forms storage gas in shale reservoirs. Shale adsorption actual reservoir not only affected by individual factors such as water content, temperature, and pressure but also synergetic effect these factors. In this study, we conducted laboratory experiments on methane dry wet at different pressures temperatures. The explored. results show that increasing temperature weakens interaction between reduces capacity due to exothermic nature adsorption. Water...
The chemistry of phosphoranyl radicals has received increasing attention in recent years. Here, we report the generation amidyl through photocatalytic deoxygenation hydroxylamines with triphenylphosphine. This methodology offers a novel and convenient route to diverse range N-acyliminophosphoranes moderate good yields under visible-light photoredox conditions. Fluorescence quenching experiments suggest that excited-state organic photocatalyst (4CzIPN) was oxidatively quenched by Cu(II) salt.
Monocular 3D object detection has become a mainstream approach in automatic driving for its easy application. A prominent advantage is that it does not need Li-DAR point clouds during the inference. However, most current methods still rely on cloud data labeling ground truths used training phase. This inconsistency between and inference makes hard to utilize large-scale feedback increases collection expenses. To bridge this gap, we propose new weakly supervised monocular objection method,...
Monocular depth estimation is known as an ill-posed task in which objects a 2D image usually do not contain sufficient information to predict their depth. Thus, it acts differently from other tasks (e.g., classification and segmentation) many ways. In this paper, we find that self-supervised monocular shows direction sensitivity environmental dependency the feature representation. But current backbones borrowed pay less attention handling different types of information, limiting overall...
In the field of autonomous driving, two important features driving car systems are explainability decision logic and accuracy environmental perception. This paper introduces DME-Driver, a new system that enhances performance reliability system. DME-Driver utilizes powerful vision language model as decision-maker planning-oriented perception control signal generator. To ensure explainable reliable decisions, logical is constructed based on large model. follows employed by experienced human...
Fluid flow in porous media is the key scientific problem development of oil and gas reservoirs. The traditional mechanics fluid which based on continuum hypothesis Darcy′s law plays an important role developing conventional resources. In recent years, unconventional reservoirs are drawing more attention all over world, therefore theory technology, especially corresponding mechanisms have become hot research issues. exhibit distinct multiscale characteristics, even with six orders magnitude...
Fusing features from different sources is a critical aspect of many computer vision tasks. Existing approaches can be roughly categorized as parameter-free or learnable operations. However, modules are limited in their ability to benefit offline learning, leading poor performance some challenging situations. Learnable fusing methods often space-consuming and timeconsuming, particularly when with shapes. To address these shortcomings, we conducted an in-depth analysis the limitations...
Methane adsorption and desorption in shale can significantly be affected by water due to the water-bearing depositional environment of application hydraulic fracturing technology gas production. The characteristics are comprehensively temperature, pressure, especially, content reservoir. To further explore impact on desorption, adsorption-desorption experiments methane at different temperatures pressures performed. Afterward, behavior hysteresis characterized employing Langmuir model...
Abstract Shale, a heterogeneous and extremely complex gas reservoir, contains low porosity ultra-Low permeability properties at different pore scales. Its flow behaviors are more complicated due to forms of regimes under laboratory conditions. Flow change with respect scale variation resulting in permeability. This work presents new insights regarding the radius adsorption, effective stress impact both on shale measurements regimes. From this study, it was revealed that value Klinkenberg...
The curve-based lane representation is a popular approach in many detection methods, as it allows for the of lanes whole object and maximizes use holistic information about lanes. However, curves produced by these methods may not fit well with irregular lines, which can lead to gaps performance compared indirect representations such segmentation-based or point-based methods. We have observed that are intended be irregular, but they appear zigzagged perspective view due being drawn on uneven...
Referring multi-object tracking (RMOT) aims at detecting and multiple objects following human instruction represented by a natural language expression. Existing RMOT benchmarks are usually formulated through manual annotations, integrated with static regulations. This approach results in dearth of notable diversity constrained scope implementation. In this work, our key idea is to bootstrap the task referring introducing discriminative words as much possible. specific, we first develop...