A5100423306- Planetary Science and Exploration
- Astro and Planetary Science
- Robotic Path Planning Algorithms
- Visual Attention and Saliency Detection
- Image and Video Quality Assessment
- Video Surveillance and Tracking Methods
- Advanced Image and Video Retrieval Techniques
- Music and Audio Processing
- High-pressure geophysics and materials
- Advanced Image Processing Techniques
- Control and Dynamics of Mobile Robots
- Video Analysis and Summarization
- Image Retrieval and Classification Techniques
- Multimodal Machine Learning Applications
- Image and Signal Denoising Methods
- Image Enhancement Techniques
- Advanced Statistical Modeling Techniques
- Video Coding and Compression Technologies
- 3D Modeling in Geospatial Applications
- Aesthetic Perception and Analysis
- Chemical Synthesis and Characterization
- Chaos-based Image/Signal Encryption
- Robotic Locomotion and Control
- Guidance and Control Systems
- Soil Mechanics and Vehicle Dynamics
Shanghai Ocean University
2024
Huazhong Agricultural University
2023
China Academy of Space Technology
2022
BOE Technology Group (China)
2022
University of California, Davis
2019
Beihang University
2017
Purdue University West Lafayette
2014
Harbin University of Commerce
2011
Dalian University of Technology
2010
Peking University
2006
Video summarization aims to distill the most important information from a source video into either an abridged clip or textual narrative. Existing methods often treat generation of and text summaries as independent tasks, thus neglecting semantic correlation between visual summarization. In other words, these only study single modality output without considering coherent outputs. this work, we first introduce novel task: cross-modal This task seeks transfer long condensed semantically...
A spherical mobile robot, BHQ-1, designed for environment exploration, was briefly introduced. The dynamic model of BHQ-1 established with a simplified Boltzmann-Hamel equation. From the expressions input moments two motors to drive robot are deduced when moves along straight trajectory and circular separately. Simulation result track complex experiments avoid an obstacle show planning method is effective.
Microwave and millimeter-wave (mm-wave)-based diagnostics have proven invaluable in the understanding of magnetic confinement fusion plasma science found wide application at major facilities, including DIII-D, EAST, ASDEX-U, HL-2A, KSTAR, J-TEXT. This trend is expected to continue unabated future burning devices such as ITER DEMO. Attractive features include their nonperturbing nature relatively modest access requirements due short wavelength, well recent dramatic advances state technology....
The click-through rate (CTR) prediction task is to predict whether a user will click on the recommended item. As mind-boggling amounts of data are produced online daily, accelerating CTR model training critical ensuring an up-to-date and reducing cost. One approach increase speed apply large batch training. However, as shown in computer vision natural language processing tasks, with easily suffers from loss accuracy. Our experiments show that previous scaling rules fail neural networks. To...
Abstract With advancements in hardware and the rapid development of internet, clarity surveillance videos has significantly improved, allowing people to quickly understand incidents a specific location without being physically present. This is especially crucial maritime navigation, where there lack solid ground feeling safety. High-definition transmitted from sea become medium for land-based entities ensure However, due absence base stations at sea, these must be via satellite...
The edge-harmonic oscillation (EHO) mode is one of the characteristic modes that provides an edge transport channel during high-confinement ELM-free phase (the so-called Quiescent High-confinement phase). Recent theoretical work and extensive experimental observations have suggested large rotational E × B shear key to destabilize EHO. As eigenmode grows amplitude, it will exert a drag on rotation, resulting in EHO's saturation (Snyder et al 2007 Nucl. Fusion 47 961). However, detailed...
Graph stream, referred to as an evolving graph with a timing sequence of updated edges through continuous is emerging data format widely used in big applications. Coping stream challenging because: 1) fully storing the continuously produced and extremely large-scale datasets difficult if not impossible; 2) supporting queries relevant both topology temporal information nontrivial. Recently, summarization techniques have attracted much attention providing approximate storage query processing...
On-line fashion communities in which participants post photos of personal items for viewing and possible purchase by others are becoming increasingly popular. Generally, these taken individuals who have no training photography with low-cost mobile phone cameras. It is desired that the products high aesthetic quality to improve users' online shopping experience. In this work, we design features inference context shopping. Psychophysical experiments conducted construct a database photos'...
According to the helicopter flight performance, a route planning method based on interfered fluid dynamic system is proposed for agricultural plant protection. The model of used determine maximum lift, climb performance and front speed. obstacle equivalent established, which quantified into disturbance matrix initial flow field corrected generate new waypoint. Combined with identified characteristics, different avoidance strategies are made types obstacles. simulation results show that...
Manned lunar mobile system is one of significant rovers in exploration project, and suspension the basis to land travel large-scale smoothly on surface. For a purpose directly landing surface under heavy-load condition, needs meet kinds special requirements achieve buffering except for general requirements. Therefore, an integrated active scheme with adjustable attitude, variable damping large stroke proposed. Double-arm structure friction damper are applied scheme, change adjustment...
This paper presented a new method for 2D plus depth stereoscopic video coding. gave the scalable coding framework based on wavelet. Proposed temporal compensated filtering (DCTF) realized scalable. The of encoder, extractor and decoder were given. experimental results showed that sequences temporal, space quality realized.
The online detection of action start in video data has witnessed an increase attention from both academia and industry, for abundant use-cases (e.g., alert mechanism videos used surveillance with ability to automate the recording key frames timestamp). Conventional approaches heavily rely on frame-level annotations other prior knowledge that can only be applied limited categories. In this paper, we introduce Generic Action Start Detection (GASD): a new task aims detect taxonomy-free manner....
Video summarization aims to distill the most important information from a source video produce either an abridged clip or textual narrative. Traditionally, different methods have been proposed depending on whether output is text, thus ignoring correlation between two semantically related tasks of visual and summarization. We propose new joint text task. The goal generate both shortened along with corresponding summary long video, collectively referred as cross-modal summary. generated...
With declining populations in the wild, captive rescue and breeding have become one of most important ways to protect pangolins from extinction. At present, success rate artificial is low, due insufficient understanding behavior characteristics pangolins. The automatic recognition method based on machine vision not only monitors for 24 hours but also reduces stress response This paper aimed establish a temporal relation attention mechanism network (Pangolin & transfer network, PBATn) monitor...
The click-through rate (CTR) prediction task is to predict whether a user will click on the recommended item. As mind-boggling amounts of data are produced online daily, accelerating CTR model training critical ensuring an up-to-date and reducing cost. One approach increase speed apply large batch training. However, as shown in computer vision natural language processing tasks, with easily suffers from loss accuracy. Our experiments show that previous scaling rules fail neural networks. To...
With the growing demand for outdoor display, a number of new display processing methods emerged, so how to reliably access quality products is greatly important in determining which better between two different products. To solve such difficulty, we this paper put forward novel subjective evaluation method, combines single stimulation method and comparison mutually supervise each other thus highly enhances reliability scoring results. We conduct newly proposed on BOE RGBW screen typical RGB...