- Microfluidic and Bio-sensing Technologies
- Electrowetting and Microfluidic Technologies
- Multimodal Machine Learning Applications
- Wireless Power Transfer Systems
- Visual Attention and Saliency Detection
- Gastrointestinal Bleeding Diagnosis and Treatment
- Advanced Neural Network Applications
- Text and Document Classification Technologies
- Orbital Angular Momentum in Optics
- Advanced Malware Detection Techniques
- Energy Harvesting in Wireless Networks
- Network Security and Intrusion Detection
- Soft Robotics and Applications
- Micro and Nano Robotics
- Anomaly Detection Techniques and Applications
- Microfluidic and Capillary Electrophoresis Applications
- Advanced Thermoelectric Materials and Devices
- Gaze Tracking and Assistive Technology
- Domain Adaptation and Few-Shot Learning
- Voice and Speech Disorders
- Acoustic Wave Resonator Technologies
- Solar Radiation and Photovoltaics
- Machine Learning and Data Classification
- Advanced ceramic materials synthesis
- Microbial Inactivation Methods
Beihang University
2020-2024
Shanxi Agricultural University
2024
Software (Spain)
2024
The University of Sydney
2020-2023
Massey University
2018-2021
Zhejiang University
2021
The simultaneous recognition of multiple objects in one image remains a challenging task, spanning events the field such as various object scales, inconsistent appearances, and confused inter-class relationships. Recent research efforts mainly resort to statistic label co-occurrences linguistic word embedding enhance unclear semantics. Different from these researches, this paper, we propose novel Transformer-based Dual Relation learning framework, constructing complementary relationships by...
Salient object detection (SOD) is a crucial and preliminary task for many computer vision applications, which have made progress with deep CNNs. Most of the existing methods mainly rely on RGB information to distinguish salient objects, faces difficulties in some complex scenarios. To solve this, recent RGBD-based networks are proposed by adopting depth map as an independent input fuse features information. Taking advantages RGBD methods, we propose novel depth-aware framework, has following...
Conventional RGB-D salient object detection methods aim to leverage depth as complementary information find the regions in both modalities. However, results heavily rely on quality of captured data which sometimes are unavailable. In this work, we make first attempt solve problem with a novel depth-awareness framework. This framework only relies RGB testing phase, utilizing supervision for representation learning. To construct our well achieving accurate results, propose Ubiquitous Target...
Abstract Single‐cell arrays have emerged as a versatile method for executing single‐cell manipulations across an array of biological applications. In this paper, innovative microfluidic platform is unveiled that utilizes optoelectronic tweezers (OETs) to and sort individual cells at flow rate 20 µL min −1 . This also adept dielectrophoresis (DEP)‐based, light‐guided retrievals from designated micro‐wells. presents compelling non‐contact the rapid straightforward sorting are hard distinguish....
Multi-label image recognition aims to recognize multiple objects simultaneously in one image. Recent ideas solve this problem have focused on learning dependencies of label co-occurrences enhance the high-level semantic representations. However, these methods usually neglect important relations intrinsic visual structures and face difficulties understanding contextual relationships. To build global scope context as well interactions between modality linguistic modality, we propose...
A currently popular research area in end-to-end speech translation is the use of knowledge distillation from a machine (MT) task to improve (ST) task. However, such scenario obviously only allows one way transfer, which limited by performance teacher model. Therefore, We hypothesis that distillation-based approaches are sub-optimal. In this paper, we propose an alternative–a trainable mutual-learning scenario, where MT and ST models collaboratively trained considered as peers, rather than...
A deep learning-assisted automated separation platform of single cells and microparticles using optoelectronic tweezers was proposed in this paper, which allows accurate manipulation long-term dynamic observation without complex microfluidic structures. single-cell detector based on YOLOv5 developed to realize automation high throughput recognition chips. Then these recognized or particles were captured operated by the light patterns generated system. We built a realized automatic queuing...
Optoelectronic Tweezers In article number 2307329, Lin Feng and co-workers constructed a novel microfluidic chip based on an optoelectronic tweezers system, comprising micro-well array integrated with photo-conductive plate. This method enables the parallel capture of cells, forming effective cell arrays in continuous flow. It also allows for light-guided single-cell retrievals by modulating electric field generating negative dielectrophoresis force target micro-well.
The untethered microrobots driven by multiple external physics fields have promising ability in minimally invasive disease treatments. One common type of the driving is gradient magnetic field, which can provide with adequate force complicated environment. In this study, a control method microrobot through field system presented, realized moving free point (FFP) to produce an alterable force. A confirmatory experiment robot reciprocating motion undertaken 1D system. could be applied further...
Endoscopy is an effective tool for early diagnosis and treatment of gastrointestinal diseases. An increasing number medical institutions prefer wireless capsule endoscopy (WCE) rather than traditional endoscopies. However, the that launched in current commercial market have disadvantages inability active locomotion low data transmission rate. This paper presents a 6 square coils electromagnetic control system with Wi-Fi based video module WCE. By adjusting driven these coils, device able to...
Photovoltaic (PV) power is an important part of renewable energy and has been distributed more widely. Due to the characteristics intermittency randomness, accurate forecasting short-term PV output will ensure secure operation economic integration in smart grids. However, previous works on show a strong dependency weather conditions common, that means there exists sophisticated relationship which hard be described quantitatively between various influencing factors power. What we have focused...
Salient object detection (SOD) is a crucial and preliminary task for many computer vision applications, which have made progress with deep CNNs. Most of the existing methods mainly rely on RGB information to distinguish salient objects, faces difficulties in some complex scenarios. To solve this, recent RGBD-based networks are proposed by adopting depth map as an independent input fuse features information. Taking advantages RGBD methods, we propose novel depth-aware framework, has following...
Magnetically controlled capsule endoscope robot has developed rapidly. A key problem of is how to precisely control its movement in human body. Thus, the position and orientation parameters should be detected real time. This paper proposes a novel localization method based on simulated annealing-particle swarm optimization (SA-PSO) algorithm. system with 4×4 Hall sensor array designed used test accuracy method. Static experimental results show that average total error 1.64 mm, 1.08°. The at...
Abstract In this paper, a simulation prototype of high-power radioisotope thermoelectric generator (RTG) based on skutterudite is designed. By replacing the isotope heat source with an electrically heated source, its output power 40 W under condition 1600 electric heating, and conversion efficiency can reach 2.5%. The contains total 8 modules, each module 16 pairs materials. Its maximum theoretical 66 temperature 550 °C. This research provides leading guide reference point for difference...
Label noise is an important part in the process of machine learning. Transition matrix provides effective way to reduce impact label on classification algorithm. In this experiment, we study logistic regression algorithm and random forest We use known real transition evaluate robustness two datasets. also design a estimator estimate three datasets algorithms. average error effectiveness top-1 accuracy our method.
For the successful and efficient operation of capsule endoscope robots in human stomach, high-accuracy multi-degree freedom (MDOF) motion control is crucial. This paper constructs a human-sized magnetic system proposes 5-DOF method that utilizes electromagnetic coils localization approach based on simulated annealing-particle swarm optimization algorithm, aiming to enhance accuracy deflection control. The results indicate proposed methods achieve excellent levitation stability improved...
Wireless capsule endoscopy (WCE) is an effective tool for gastrointestinal diseases diagnosis. Compare to the traditional endoscopy, more and patients medical institutions prefer WCE due its user-friendly non-cross infection. However, size limitation of would lead a problem insufficient energy. This paper will propose relay-based power allocation algorithm transmission system based on market equilibrium scheme. Considering resources as assigned networks revenue defined by logarithm utility...
This work presents a novel electrical method, implemented in the form of microfluidic device, for cell arraying and target lysis. The device contains micro-well array on photoconductive layer based optoelectronic tweezers (OET) where parallel manipulation is performed. As suspension flows over micro-wells, cells can be actively captured micro-wells by light-induced dielectrophoresis (DEP) forces, designed pattern less than 120 s. single-cell capture rate 83 % patterned array, about 94% are...
Optoelectronic tweezers (OETs) based on dielectrophoresis (DEP) force is a valuable tool for the manipulation of particles and cells. However, DEP-based methods that can measure electrical parameters are always preformed static metal electrode DEP systems. Here, we present partitioned single-sided OET chip combines an system microfluidic channel. Unlike classical sandwich-structure chip, close to but switch functions easily. Numerical simulations studied analyze electric field provide data...