A5000564739- Human Mobility and Location-Based Analysis
- Traffic Prediction and Management Techniques
- Video Surveillance and Tracking Methods
- Anomaly Detection Techniques and Applications
- Time Series Analysis and Forecasting
- Autonomous Vehicle Technology and Safety
- Data Management and Algorithms
- Transportation Planning and Optimization
- Context-Aware Activity Recognition Systems
- Advanced Image and Video Retrieval Techniques
- Topic Modeling
- Indoor and Outdoor Localization Technologies
- Energy Load and Power Forecasting
- Spam and Phishing Detection
- Recommender Systems and Techniques
- Infrared Target Detection Methodologies
- Speech and Audio Processing
- Power Line Communications and Noise
- Music and Audio Processing
- Cryptography and Data Security
- Domain Adaptation and Few-Shot Learning
- Building Energy and Comfort Optimization
- Atmospheric and Environmental Gas Dynamics
- Privacy-Preserving Technologies in Data
- Data-Driven Disease Surveillance
UNSW Sydney
2022-2025
First Affiliated Hospital of Anhui Medical University
2025
Anhui Medical University
2025
Wuhan Polytechnic University
2025
Yantai University
2024
Chinese Academy of Sciences
2017-2024
Shenyang Research Institute of Foundry
2024
Xidian University
2024
Changchun Institute of Optics, Fine Mechanics and Physics
2017-2024
Gansu Agricultural University
2024
Pedestrian trajectory prediction is an extremely challenging problem because of the crowdedness and clutter scenes. Previous deep learning LSTM-based approaches focus on neighbourhood influence pedestrians but ignore scene layouts in pedestrian prediction. In this paper, a novel hierarchical network proposed to consider both social layouts. Our SS-LSTM, which stands for Social-Scene-LSTM, uses three different LSTMs capture person, scale information. We also use circular shape setting instead...
Generative Adversarial Networks (GANs) have shown remarkable success in producing realistic-looking images the computer vision area. Recently, GAN-based techniques are to be promising for spatio-temporal-based applications such as trajectory prediction, events generation, and time-series data imputation. While several reviews GANs been presented, no one has considered addressing practical challenges relevant spatio-temporal data. In this article, we conducted a comprehensive review of recent...
This paper presents a new perspective on time series forecasting. In existing forecasting methods, the models take sequence of numerical values as input and yield output. The SOTA are largely based Transformer architecture, modified with multiple encoding mechanisms to incorporate context semantics around historical data. Inspired by successes pre-trained language foundation models, we pose question about whether these can also be adapted solve time-series Thus, propose paradigm:...
Traditional Light Detection and Rangings (LiDARs) can quickly collect high-accuracy of three-dimensional (3D) point cloud data at a designated wavelength (i.e., cannot obtain hyperspectral data), while the passive imager rich spectral ground objects, but are lack 3D spatial data. This paper presents one innovative study on design airborne-oriented supercontinuum laser (SCLaHS) LiDAR with 50 bands covering 400 nm to 900 resolution 10 sampling distance (GSD) 0.5 m. The major innovations...
Self-Supervised Learning (SSL) is a new paradigm for learning discriminative representations without labeled data, and has reached comparable or even state-of-the-art results in comparison to supervised counterparts. Contrastive (CL) one of the most well-known approaches SSL that attempts learn general, informative data. CL methods have been mostly developed applications computer vision natural language processing where only single sensor modality used. A majority pervasive computing...
In this paper, we propose a novel pipeline that leverages language foundation models for temporal sequential pattern mining, such as human mobility forecasting tasks. For example, in the task of predicting Place-of-Interest (POI) customer flows, typically number visits is extracted from historical logs, and only numerical data are used to predict visitor flows. research, perform directly on natural input includes all kinds information values contextual semantic information. Specific prompts...
As a decisive part in the success of Mobility-as-a-Service (MaaS), spatio-temporal predictive modeling for crowd movements is challenging task particularly considering scenarios where societal events drive mobility behavior deviated from normality. While tremendous progress has been made to model high-level regularities with deep learning, most, if not all existing methods are neither aware dynamic interactions among multiple transport modes nor adaptive unprecedented volatility brought by...
The coronavirus disease 2019 (COVID-19) pandemic has induced a significant global concern on mental health. However few studies have measured the ability of individuals to "withstand setbacks, adapt positively, and bounce back from adversity" scale. We aimed examine level resilience, its determinants, association with maladaptive coping behaviours during pandemic.The Association Pacific Rim Universities (APRU) conducted survey involving 26 countries by online, self-administered questionnaire...
Trajectory prediction is an important task to support safe and intelligent behaviours in autonomous systems. Many advanced approaches have been proposed over the years with improved spatial temporal feature extraction. However, human behaviour naturally multimodal uncertain: given past trajectory surrounding environment information, agent can multiple plausible trajectories future. To tackle this problem, essential named (MTP) has recently studied, which aims generate a diverse, acceptable...
The next Point of Interest (POI) recommendation task is to predict users' immediate POI visit given their historical data. Location-Based Social Network (LBSN) data, which often used for the task, comes with challenges. One frequently disregarded challenge how effectively use abundant contextual information present in LBSN Previous methods are limited by numerical nature and fail address this challenge. In paper, we propose a framework that uses pretrained Large Language Models (LLMs) tackle...
Pedestrian trajectory prediction is important in various applications such as driverless vehicles, social robots, intelligent tracking systems and space planning. Existing methods focus on analysing the influence of neighbours but ignore effect intended destinations pedestrians which also plays a key role route In this paper, we propose novel two- stage method to yield multiple trajectories with different probabilities towards destination regions scene. Our method, refer Bi-Prediction, uses...
Pedestrian trajectory prediction is fundamental to a wide range of scientific research work and industrial applications. Most the current advanced methods incorporate context information such as pedestrian neighbourhood, labelled static obstacles, background scene into process. In contrast these which require rich contexts, method in our paper focuses on predicting pedestrian's future using his/her observed part only. Our method, we refer LVTA, Location-Velocity-Temporal Attention LSTM model...
Human mobility prediction is a core functionality in many location-based services and applications. However, due to the sparsity of data, it not an easy task predict future POIs (place-of-interests) that are going be visited. In this paper, we propose MobTCast, Transformer-based context-aware network for prediction. Specifically, explore influence four types context prediction: temporal, semantic, social geographical contexts. We first design base feature extractor using Transformer...
Food fraud is widespread in the aquatic food market, hence fast and non-destructive methods of identification fish flesh are needed. In this study, multispectral imaging (MSI) was used to screen slices from 20 edible species commonly found sea around Yantai, China, by combining based on mitochondrial COI gene. We that nCDA images transformed MSI data showed significant differences splices species. then employed eight models compare their prediction performances hold-out method with 70%...
Pedestrian path forecasting is crucial in applications such as smart video surveillance. It a challenging task because of the complex crowd movement patterns scenes. Most existing state-of-the-art LSTM based prediction methods require rich context like labelled static obstacles, entrance/exit regions and even background scene. Furthermore, incorporating contextual information into trajectory increases computational overhead decreases generalization models across different In this paper, we...
The usage of smartphone-collected respiratory sound, trained with deep learning models, for detecting and classifying COVID-19 becomes popular recently. It removes the need in-person testing procedures especially rural regions where related medical supplies, experienced workers, equipment are limited. However, existing sound-based diagnostic approaches in a fully supervised manner, which requires large scale well-labelled data. is critical to discover new methods leverage unlabelled data,...
Traffic forecasting is a critical task to extract values from cyber-physical infrastructures, which the backbone of smart transportation. However owing external contexts, dynamics at each sensor are unique. For example, afternoon peaks sensors near schools more likely occur earlier than those residential areas. In this paper, we first analyze real-world traffic data show that has unique dynamic. Further analysis also shows pair Then, explore how node embedding learns every location. Next,...