A5100340299- Video Surveillance and Tracking Methods
- Human Pose and Action Recognition
- Advanced Neural Network Applications
- Advanced Image and Video Retrieval Techniques
- Nanoplatforms for cancer theranostics
- Autonomous Vehicle Technology and Safety
- Distributed systems and fault tolerance
- Multimodal Machine Learning Applications
- Gait Recognition and Analysis
- Image Retrieval and Classification Techniques
- Medical Imaging and Analysis
- Handwritten Text Recognition Techniques
- Domain Adaptation and Few-Shot Learning
- RNA Interference and Gene Delivery
- Advanced Data Storage Technologies
- Cloud Computing and Resource Management
- Anomaly Detection Techniques and Applications
- Advanced biosensing and bioanalysis techniques
- Aerodynamics and Fluid Dynamics Research
- Brain Tumor Detection and Classification
- Software System Performance and Reliability
- Medical Image Segmentation Techniques
- Automated Road and Building Extraction
- Caching and Content Delivery
- Generative Adversarial Networks and Image Synthesis
The University of Sydney
2025
Soochow University
2020-2024
Liaocheng University
2024
Purdue University West Lafayette
2023
Fudan University
2023
Beijing University of Civil Engineering and Architecture
2023
Tokyo Metropolitan University
2022-2023
Sichuan University
2023
Beijing University of Chemical Technology
2023
Alibaba Group (China)
2022-2023
Relationships among objects play a crucial role in image understanding. Despite the great success of deep learning techniques recognizing individual objects, reasoning about relationships remains challenging task. Previous methods often treat this as classification problem, considering each type relationship (e.g. ride) or distinct visual phrase person-ride-horse) category. Such approaches are faced with significant difficulties caused by high diversity appearance for kind large number...
Recently, deep learning-based cross-view gait recognition has become popular owing to the strong capacity of convolutional neural networks (CNNs). Current learning methods often rely on loss functions used widely in task face recognition, e.g., contrastive and triplet loss. These have problem hard negative mining. In this paper, a robust, effective, gait-related function, called angle center (ACL), is proposed learn discriminative features. The function robust different local parts temporal...
The rapid advances of transportation infrastructure have led to a dramatic increase in the demand for smart systems capable monitoring traffic and street safety. Fundamental these applications are community-based evaluation platform benchmark object detection multi-object tracking. To this end, we organize AVSS2017 Challenge on Advanced Traffic Monitoring, conjunction with International Workshop Street Surveillance Safety Security (IWT4S), evaluate state-of-the-art tracking algorithms...
This paper introduces our solution for the Track2 in AI City Challenge 2020 (AICITY20). The is a vehicle re-identification (ReID) task with both real-world data and synthetic data.Our based on strong baseline bag of tricks (BoT-BS) proposed person ReID. At first, we propose multi-domain learning method to joint train model. Then, Identity Mining automatically generate pseudo labels part testing data, which better than k-means clustering. tracklet-level re-ranking strategy weighted features...
Multi-Target Multi-Camera Tracking has a wide range of applications and is the basis for many advanced inferences predictions. This paper describes our solution to Track 3 multi-camera vehicle tracking task in 2021 AI City Challenge (AICITY21). proposes multi-target framework guided by crossroad zones. The includes: (1) Use mature detection re-identification models extract targets appearance features. (2) modified JDE-Tracker (without module) track single-camera vehicles generate tracklets....
Improving the enrichment of drugs or theranostic agents within tumors is very vital to achieve effective cancer diagnosis and therapy while greatly reducing dosage damage normal tissues. Herein, as a proof concept, we for first time report red light-initiated probe-RNA cross-linking (RLIPRC) strategy that can not only robustly promote accumulation retention probe in tumor prolonged imaging but also significantly inhibits growth. A near-infrared (NIR) fluorescent f-CR consisting NIR dye...
Image-text retrieval is a central problem for understanding the semantic relationship between vision and language, serves as basis various visual language tasks. Most previous works either simply learn coarse-grained representations of overall image text, or elaborately establish correspondence regions pixels text words. However, close relations coarse- fine-grained each modality are important image-text but almost neglected. As result, such inevitably suffer from low accuracy heavy...
This paper introduces our solution for the Track2 in AI City Challenge 2021 (AICITY21). The is a vehicle re-identification (ReID) task with both real-world data and synthetic data. We mainly focus on four points, i.e. training data, unsupervised domain-adaptive (UDA) training, post-processing, model ensembling this challenge. (1) Both cropping using can help learn more discriminative features. (2) Since there new scenario test set that dose not appear set, UDA methods perform well (3)...
This paper addresses the challenges of creating efficient and high-quality datasets for machine learning potential functions. We present a novel approach, termed DV-LAE (Difference Vectors based on Local Atomic Environments), which utilizes properties atomic local environments employs histogram statistics to generate difference vectors. technique facilitates dataset screening optimization, effectively minimizing redundancy while maintaining data diversity. have validated optimized in...
The Environmental Impact Evaluation (EIE) of petrochemical construction projects is a critical process to evaluate potential environmental effects caused by project activities. EIE inherently complex multi-attribute decision-making (MADM) problem due the need balance these diverse factors. To address this complexity, study applies triangular fuzzy neutrosophic number cross-entropy (TFNN-CE) approach, which operates within framework sets (TFNSs). TFNSs provide robust mathematical tool handle...
Visual retrieval tasks such as image and person re-identification (Re-ID) aim at effectively thoroughly searching images with similar content or the same identity. After obtaining retrieved examples, re-ranking is a widely adopted post-processing step to reorder improve initial results by making use of contextual information from semantically neighboring samples. Prevailing approaches update distance metrics mostly rely on inefficient crosscheck set comparison operations while computing...
Nowadays, deep learning is widely applied to extract features for similarity computation in person re-identification (re-ID). However, the difference between training data and testing makes performance of learned feature degraded during testing. Hence, re-ranking proposed mitigate this issue various algorithms have been developed. most existing methods focus on replacing Euclidean distance with sophisticated metrics, which are not friendly downstream tasks hard be used fast retrieval massive...
Radiotherapy (RT)-induced in situ vaccination greatly promotes the development of personalized cancer vaccines owing to massive release antigens initiated by tumor-localized RT eliciting tumor-specific immune response. However, its broad application treatment is seriously impeded poor antigen cross-presentation, low response rate, and short duration efficacy. Herein, tumor-antigen-capturing nanosystem dAuNPs@CpG consisting gold nanoparticles, 3,5-cyclohexanedione (CHD), immunoadjuvant CpG...
Abstract Research on the time series classification is gaining an increased attention in machine learning and data mining areas due to existence of almost everywhere, especially our daily work life. Recent studies have shown that convolutional neural networks (CNN) can extract good features from images texts, but it often encounters problem low accuracy, when directly employed solve classification. In this pursuit, present study envisaged a novel combined model based slide relative position...
We extend the classical tracking-by-detection paradigm to this tracking-any-object task. Solid detection results are first extracted from TAO dataset. Some state-of-the-art techniques like \textbf{BA}lanced-\textbf{G}roup \textbf{S}oftmax (\textbf{BAGS}\cite{li2020overcoming}) and DetectoRS\cite{qiao2020detectors} integrated during detection. Then we learned appearance features represent any object by training feature learning networks. ensemble several models for improving representation....
Face clustering has attracted rising research interest recently to take advantage of massive amounts face images on the web. State-of-the-art performance been achieved by Graph Convolutional Networks (GCN) due their powerful representation capacity. However, existing GCN-based methods build graphs mainly according kNN relations in feature space, which may lead a lot noise edges connecting two faces different classes. The features will be polluted when messages pass along these edges, thus...
This paper explores the importance of detection and appearance features for multiple object tracking. Extensive detectors including hand-crafted methods deep learning have been tested. We found in this that simply improving performance can lead to much better tracking results. The data association used are Kalman Filter Hungarian algorithm as proposed [1]. CNN color histogram extracted measure similarities between objects. Our experiments show help with association. then combine together an...
Oil wells play an important role in the extraction of oil and gas, their future potential extends beyond gas exploitation to include development geothermal resources for sustainable power generation. Identifying detecting are paramount importance given crucial well distribution energy planning. In recent years, significant progress has been made single objects, with recognition accuracy exceeding 90%. However, there still remaining challenges, particularly regard small-scale varying viewing...
We present a multi-task deep learning framework to improve the performance of Multiple Object Tracking (MOT) problem. Motion and appearance cues are ombined together build an online multiple object tracker. While being accurate, our tracker also runs fast enough. have made two major contributions in this paper: (1) Learn features offline with triplet loss. (2) Train quality-aware network by sharing convolutional features. The proposed achieves state-of-art on UA-DETRAC dataset [17] while...