A5101616088- Video Surveillance and Tracking Methods
- Medical Imaging Techniques and Applications
- Advanced Image and Video Retrieval Techniques
- Advanced X-ray and CT Imaging
- Advanced Neural Network Applications
- Image Enhancement Techniques
- Energy Efficient Wireless Sensor Networks
- Medical Image Segmentation Techniques
- Image Processing Techniques and Applications
- Multimodal Machine Learning Applications
- Mobile Agent-Based Network Management
- Advanced Image Processing Techniques
- Fire Detection and Safety Systems
- Image and Video Stabilization
- Image Retrieval and Classification Techniques
- Human Pose and Action Recognition
- Image and Signal Denoising Methods
- Anomaly Detection Techniques and Applications
- Network Traffic and Congestion Control
- Domain Adaptation and Few-Shot Learning
- Sparse and Compressive Sensing Techniques
- Infrared Target Detection Methodologies
- Image Processing and 3D Reconstruction
- Advanced MRI Techniques and Applications
- Wireless Communication Networks Research
Harvard University
2025
Dalian University of Technology
2019-2024
Jiangsu University of Science and Technology
2024
PLA Information Engineering University
2010-2024
Sichuan University
2024
University of Electronic Science and Technology of China
2009-2023
Chinese PLA General Hospital
2019-2022
Huzhou University
2021
Microsoft Research Asia (China)
2021
System Equipment (China)
2019
Correlation acts as a critical role in the tracking field, especially recent popular Siamese-based trackers. The correlation operation is simple fusion manner to consider similarity between template and search region. However, itself local linear matching process, leading lose semantic information fall into optimum easily, which may be bottleneck of designing high-accuracy algorithms. Is there any better feature method than correlation? To address this issue, inspired by Transformer, work...
In this paper, we present a new tracking architecture with an encoder-decoder transformer as the key component. The encoder models global spatio-temporal feature dependencies between target objects and search regions, while decoder learns query embedding to predict spatial positions of objects. Our method casts object direct bounding box prediction problem, without using any proposals or predefined anchors. With transformer, just uses simple fully-convolutional network, which estimates...
The Visual Object Tracking challenge VOT2019 is the seventh annual tracker benchmarking activity organized by VOT initiative. Results of 81 trackers are presented; many state-of-the-art published at major computer vision conferences or in journals recent years. evaluation included standard and other popular methodologies for short-term tracking analysis as well methodology long-term analysis. was composed five challenges focusing on different domains: (i) VOTST2019 focused RGB, (ii)...
Compared with traditional short-term tracking, long-term tracking poses more challenges and is much closer to realistic applications. However, few works have been done their performance also limited. In this work, we present a novel robust real-time framework based on the proposed skimming perusal modules. The module consists of an effective bounding box regressor generate series candidate proposals target verifier infer optimal its confidence score. Based score, our tracker determines...
Visual object tracking aims to precisely estimate the bounding box for given target, which is a challenging problem due factors such as deformation and occlusion. Many recent trackers adopt multiple-stage strategy improve estimation. These methods first coarsely locate target then refine initial prediction in following stages. However, existing approaches still suffer from limited precision, coupling of different stages severely restricts method's transferability. This work proposes novel,...
Object tracking has achieved significant progress over the past few years. However, state-of-the-art trackers become increasingly heavy and expensive, which limits their deployments in resource-constrained applications. In this work, we present LightTrack, uses neural architecture search (NAS) to design more lightweight efficient object trackers. Comprehensive experiments show that our LightTrack is effective. It can find achieve superior performance compared handcrafted SOTA trackers, such...
All instance perception tasks aim at finding certain objects specified by some queries such as category names, language expressions, and target annotations, but this complete field has been split into multiple independent sub-tasks. In work, we present a universal model of the next generation, termed UNINEXT. UNINEXT reformulates diverse unified object discovery retrieval paradigm can flexibly perceive different types simply changing input prompts. This formulation brings following benefits:...
Multi-Object Tracking (MOT) aims to detect and associate all desired objects across frames. Most methods accomplish the task by explicitly or implicitly leveraging strong cues (i.e., spatial appearance information), which exhibit powerful instance-level discrimination. However, when object occlusion clustering occur, information will become ambiguous simultaneously due high overlap among objects. In this paper, we demonstrate long-standing challenge in MOT can be efficiently effectively...
Adversarial attack of CNN aims at deceiving models to misbehave by adding imperceptible perturbations images. This feature facilitates understand neural networks deeply and improve the robustness deep learning models. Although several works have focused on attacking image classifiers object detectors, an effective efficient method for single trackers any target in a model-free way remains lacking. In this paper, cooling-shrinking is proposed deceive state-of-the-art SiameseRPN-based...
Specular reflections (i.e., highlight) always exist in endoscopic images, and they can severely disturb surgeons' observation judgment. In an augmented reality (AR)-based surgery navigation system, the highlight may also lead to failure of feature extraction or registration. this paper, we propose adaptive robust principal component analysis (Adaptive-RPCA) method remove specular image sequences. It iteratively optimize sparse part parameter during RPCA decomposition. new approach, first...
Compared with short-term tracking, long-term tracking remains a challenging task that usually requires the algorithm to track targets within local region and re-detect over entire image. However, few works have been done their performances also limited. In this paper, we present novel robust real-time framework based on proposed search module re-detection module. The consists of an effective bounding box regressor generate series candidate proposals target verifier infer optimal its...
Unmanned aerial vehicles (UAVs) equipped with high definition (HD) cameras can obtain a large number of detailed inspection images. The insulator is an indispensable component in the transmission lines. Detecting image video quickly and accurately provide reliable basis for ranging obstacle avoidance flight UAV close to tower line. At same time, serious threat safety power grid due multiple faults insulator, computer technology should be fully utilized diagnose fault. Detection images...
Compared with the conventional 1×1 acquisition mode of projection in computed tomography (CT) image reconstruction, 2×2 improves collection efficiency and reduces x-ray exposure time. However, collected based on has low resolution (LR) reconstructed quality is poor, thus limiting use this CT imaging systems. In study, a novel sinogram-super-resolution (SR) generative adversarial network model proposed to obtain high-resolution (HR) sinograms from LR sinograms, thereby improving...
Compared with traditional short-term tracking, long-term tracking poses more challenges and is much closer to realistic applications. However, few works have been done their performance also limited. In this work, we present a novel robust real-time framework based on the proposed skimming perusal modules. The module consists of an effective bounding box regressor generate series candidate proposals target verifier infer optimal its confidence score. Based score, our tracker determines...
The reference-based object segmentation tasks, namely referring image (RIS), video (RVOS), and (VOS), aim to segment a specific by utilizing either language or annotated masks as references. Despite significant progress in each respective field, current methods are task-specifically designed developed different directions, which hinders the activation of multi-task capabilities for these tasks. In this work, we end fragmented situation propose UniRef unify three tasks with single...
The intelligent transportation system (ITS) is envisioned by linking existing and emerging technologies of computers, wireless radio communications systems sophisticated sensors to be used in vehicles roads. IEEE 802.15.4 a new standard designed for low rate personal area networks (LR_WPANs) with focus on enabling sensor vehicular, residential, commercial industrial applications. This characterized its simplicity, data rate, power consumption cost networks. release anticipated make...
Multi-Object Tracking (MOT) aims to detect and associate all desired objects across frames. Most methods accomplish the task by explicitly or implicitly leveraging strong cues (i.e., spatial appearance information), which exhibit powerful instance-level discrimination. However, when object occlusion clustering occur, information will become ambiguous simultaneously due high overlap among objects. In this paper, we demonstrate long-standing challenge in MOT can be efficiently effectively...
Extracting distinctive scale invariant features from images of the same scene or object is very important in many computer vision applications, and there has been significant research into feature detectors descriptors. Some these methods have emphasized on computational speed accuracy, so that they can enable lots real-time applications with reduced requirements better performance. The purpose this paper to introduce a new octagonal center-surround detector, named OCT, give modified SURF...