A5101966265- Advanced Vision and Imaging
- Robotics and Sensor-Based Localization
- Optical measurement and interference techniques
- Advanced DC-DC Converters
- 3D Shape Modeling and Analysis
- Microgrid Control and Optimization
- Multilevel Inverters and Converters
- Human Pose and Action Recognition
- Image Processing Techniques and Applications
- Human Motion and Animation
- Advanced Sensor and Energy Harvesting Materials
- Scientific Measurement and Uncertainty Evaluation
- 3D Surveying and Cultural Heritage
- Power Systems and Renewable Energy
- Image Processing and 3D Reconstruction
- Advanced Image and Video Retrieval Techniques
- Chemical and Environmental Engineering Research
- High-Voltage Power Transmission Systems
- Hydrology and Watershed Management Studies
- Conducting polymers and applications
- Computer Graphics and Visualization Techniques
- Fish Ecology and Management Studies
- Video Analysis and Summarization
- Speech and dialogue systems
- Industrial Vision Systems and Defect Detection
Alibaba Group (China)
2024
Alibaba Group (United States)
2024
Central South University
2024
Hohai University
2021
Wuhan University of Technology
2021
University of Oklahoma
1998
Estimating the accurate depth from a single image is challenging since it inherently ambiguous and ill-posed. While recent works design increasingly complicated powerful networks to directly regress map, we take path of CRFs optimization. Due expensive computation, are usually performed between neighborhoods rather than whole graph. To leverage potential fully-connected CRFs, split input into windows perform FC-CRFs optimization within each window, which reduces computation complexity makes...
There is an emerging trend of using neural implicit functions for map representation in Simultaneous Localization and Mapping (SLAM). Some pioneer works have achieved encouraging results on RGB-D SLAM. In this paper, we present a dense RGB SLAM method with representation. To reach challenging goal without depth input, introduce hierarchical feature volume to facilitate the decoder. This design effectively fuses shape cues across different scales reconstruction. Our simultaneously solves...
DC-link voltage pulsation is a technical challenge for grid-connected three-phase source rectifiers (VSRs) under unbalanced grids. To suppress ripples arising from grids, frequency-divided resistance-emulating control strategy proposed in this paper. According to the idea, loop consists of two parts: one dedicated regulating average and other focused on suppressing pulsating voltage. The inner synchronized based approach. offers following advantages: 1) grid sensors are eliminated; 2)...
Owing to the direct coupling between input and output, input/output currents of matrix converter (DMC) are sensitive grid disturbances, especially unbalanced grids. To mitigate adverse effects caused by grids, an algebraic modulation strategy is developed for DMC in this article. Based on theory, general solution vector first derived. By utilizing free variables, sinusoidal ensured despite whether balanced or not. The impact offset variables analyzed they designed realize three-level output...
The increase in the rate of water renewal driven by hydrodynamics contributes to improving quality plain river network. Taking lakeside network Wuxi as an example, through numerical simulation, polynomial fitting, correlation analysis, and principal component hydrodynamic responses urban lake-connected networks diversion grouping were researched. Based on model influence weight we explored improvement conditions with strong human intervention high algal diversion. results showed that: (1)...
Monocular scene reconstruction from posed images is challenging due to the complexity of a large environment. Recent volumetric methods learn directly predict TSDF volume and have demonstrated promising results in this task. However, most focus on how extract fuse 2D features 3D feature volume, but none them improve way aggregated. In work, we propose an SDF transformer network, which replaces role CNN for better aggregation. To reduce explosive computation multi-head attention, sparse...
Generalizable 3D object reconstruction from single-view RGB-D images remains a challenging task, particularly with real-world data. Current state-of-the-art methods develop Transformer-based implicit field learning, necessitating an intensive learning paradigm that requires dense query-supervision uniformly sampled throughout the entire space. We propose novel approach, IPoD, which harmonizes point diffusion. This approach treats query points for as noisy cloud iterative denoising, allowing...
Generating motion sequences conforming to a target style while adhering the given content prompts requires accommodating both and style. In existing methods, information usually only flows from content, which may cause conflict between harming integration. Differently, in this work we build bidirectional control flow also adjusting towards case style-content collision is alleviated dynamics of better preserved Moreover, extend stylized generation one modality, i.e. motion, multiple...
There are increasing interests of studying the video-to-depth (V2D) problem with machine learning techniques. While earlier methods directly learn a mapping from images to depth maps and camera poses, more recent works enforce multi-view geometry constraints through optimization embedded in framework. This paper presents novel method based on recurrent neural networks further exploit potential V2D. Specifically, our optimizer alternately updates poses iterations minimize feature-metric cost,...
3D motion estimation including scene flow and point cloud registration has drawn increasing interest. Inspired by 2D estimation, recent methods employ deep neural networks to construct the cost volume for estimating accurate flow. However, these are limited fact that it is difficult define a search window on clouds because of irregular data structure. In this paper, we avoid irregularity simple yet effective method.We decompose problem into two interlaced stages, where flows optimized...
The system is an AC sine wave online uninterruptible power supply designed with STM32F407 as the main controller and full-bridge single-phase inverter core technology. based on single-chip microcomputer, consists of a rectifier circuit, BOOST PFC factor correction circuit. When input switched between 29V~43V 24V DC supply, output voltage can maintain constant frequency amplitude. Experimental data shows that has high operating efficiency excellent stability.