A5074747567- Video Surveillance and Tracking Methods
- Human Pose and Action Recognition
- Advanced Image and Video Retrieval Techniques
- Advanced Neural Network Applications
- Video Coding and Compression Technologies
- Real-time simulation and control systems
- Systemic Sclerosis and Related Diseases
- Color Science and Applications
- 3D Shape Modeling and Analysis
- Infrared Target Detection Methodologies
- Advanced Differential Equations and Dynamical Systems
- Advanced Vision and Imaging
- Image Enhancement Techniques
- Radiomics and Machine Learning in Medical Imaging
- Optical measurement and interference techniques
- Image Processing Techniques and Applications
- Mathematical and Theoretical Epidemiology and Ecology Models
- Software Reliability and Analysis Research
- Domain Adaptation and Few-Shot Learning
- CCD and CMOS Imaging Sensors
- Graph theory and applications
- Advanced Banach Space Theory
- Ultrasound Imaging and Elastography
- Reliability and Maintenance Optimization
- Calibration and Measurement Techniques
Tampere University
2019-2023
Bohai University
2008-2021
Liaoning Normal University
2020
Shandong Normal University
2020
Shandong University of Technology
2019
Harvard University
2018
South China University of Technology
2018
Brigham and Women's Hospital
2018
Chinese Academy of Medical Sciences & Peking Union Medical College
2017
University of California, San Diego
2017
In the current object detection field, one of fastest algorithms is Single Shot Multi-Box Detector (SSD), which uses a single convolutional neural network to detect in an image. Although SSD fast, there big gap compared with state-of-the-art on mAP. this paper, we propose method improve algorithm increase its classification accuracy without affecting speed. We adopt Inception block replace extra layers SSD, and call (I-SSD). The proposed can catch more information increasing complexity....
The Visual Object Tracking challenge VOT2021 is the ninth annual tracker benchmarking activity organized by VOT initiative. Results of 71 trackers are presented; many state-of-the-art published at major computer vision conferences or in journals recent years. was composed four sub-challenges focusing on different tracking domains: (i) VOT-ST2021 focused short-term RGB, (ii) VOT-RT2021 "real-time" (iii) VOT-LT2021 long-term tracking, namely coping with target disappearance and reappearance...
RGBD (RGB plus depth) object tracking is gaining momentum as sensors have become popular in many application fields such robotics. However, the best trackers are extensions of state-of-the-art deep RGB trackers. They trained with data and depth channel used a sidekick for subtleties occlusion detection. This can be explained by fact that there no sufficiently large datasets to 1) train "deep trackers" 2) challenge sequences which cue essential. work introduces new dataset - Depth-Track has...
Abstract We propose a full processing pipeline to acquire anthropometric measurements from 3D measurements. The first stage of our is commercial point cloud scanner. In the second stage, pre-defined body model fitted captured cloud. have generated one male and female SMPL library. fitting process based on non-rigid iterative closest algorithm that minimizes overall energy distance local stiffness terms. third we measure multiple circumference paths surface use nonlinear regressor provide...
To assess the value of ultrasonography (US) features for determining malignant potential complex cystic lesions.Seventy-nine lesions were reviewed retrospectively. They classified into four types according to US in type I, masses have a thick outer wall, internal septa, or both; II, are an intracystic with one more discrete solid mural within cyst; III, contain mixed and components at least 50% portion mass; IV, there predominantly (at 50%) eccentric central foci. Positive predictive values...
Accurate estimation of anthropometric body measurements from RGB images has many potential applications in industrial design, online clothing, medical diagnosis and ergonomics. Research on this topic is limited by the fact that there exist only generated datasets which are based fitting a 3D mesh to scans commercial CAESAR dataset. For 2D silhouettes generated. To circumvent data bottleneck, we introduce new scan dataset 2,675 female 1,474 male scans. We also small 200 tape measured ground...
Transfer learning methods have demonstrated state-of-the-art performance on various small-scale image classification tasks. This is generally achieved by exploiting the information from an ImageNet convolution neural network (ImageNet CNN). However, transferred CNN model with high computational complexity and storage requirement. It raises issue for real-world applications, especially some portable devices like phones tablets without high-performance GPUs. Several approximation been proposed...
Anthropometric body measurements are important for industrial design, garment fitting, medical diagnosis and ergonomics. A number of methods have been proposed to estimate the from images, but progress has slow due lack realistic publicly available datasets. The existing works train test on silhouettes 3D meshes obtained by fitting a human model commercial CAESAR scans. In this work, we introduce BODY-fit dataset that contains fitted 2,675 female 1,474 male We unify evaluation CAESAR-fit...
In order to obtain the accurate exposure time in real-time for an on-board camera different urban environments, proposed algorithm divides captured image into 5×5 sub-areas, calculates each sub-area's average brightness value get a histogram, analyzes peak distribution histogram determine what environments automobile is in. According environment works in, includes two modes: normal lit condition and high-contrast condition. It executes appropriate adjustment mechanism modes by analyzing...
In the real world, a scene is usually cast by multiple illuminants and herein we address problem of spatial illumination estimation. Our solution based on detecting gray pixels with help flash photography. We show that photography significantly improves performance pixel detection without illuminant prior, training data or calibration flash. also introduce novel dataset generated from MIT intrinsic dataset.
In the real world, a scene is usually cast by multiple illuminants and herein we address problem of spatial illumination estimation. Our solution based on detecting gray pixels with help flash photography. We show that photography significantly improves performance pixel detection without illuminant prior, training data or calibration flash. also introduce novel dataset generated from MIT intrinsic dataset.
Redundancy techniques are widely used to design fault tolerant systems. Diversity has long been protect redundant systems from Common Cause Failure (CCF). Whilst there is clear evidence that diversity can bring benefits, these benefits be difficult quantify. Therefore, a novel method which researches the effect of on CCF mitigation component degradation state point view proposed in this paper. Four categories impact factors (IFs), influence failure behavior component, defined and...
Abstract Haptic human-computer interaction technology is an important part of in virtual reality. It simulates the process human’s haptic perception real objects and feeds back information environment to people, greatly improving interactivity telepresence environment. Starting from differences between expressions, this paper summarizes research status modeling rendering equipment, analyzes characteristics various modes. The framework, progress application sensing system based on cloud are...
It is important to consider diagnostic coverage (DC) of the automatic diagnosis (AD) scheme in course design highly reliable systems which redundancy technique usually applied. Considering nature AD introduce some form redundancy, its performance will be affected by common cause failure (CCF). Due that diversity an effective antidote for CCF, it desirable assess effect and CCF on DC. Therefore, a novel method proposed quantitatively compare two different schemes include identical diverse...