A5070256347- Human Pose and Action Recognition
- Anomaly Detection Techniques and Applications
- Environmental Impact and Sustainability
- Hand Gesture Recognition Systems
- Domain Adaptation and Few-Shot Learning
- Video Surveillance and Tracking Methods
- Gait Recognition and Analysis
- Digital Media Forensic Detection
- Adversarial Robustness in Machine Learning
- Remote Sensing and LiDAR Applications
- Air Quality and Health Impacts
- Time Series Analysis and Forecasting
- Energy, Environment, Economic Growth
- Advanced Vision and Imaging
- Sustainability and Ecological Systems Analysis
- Generative Adversarial Networks and Image Synthesis
- Advanced Neural Network Applications
- COVID-19 diagnosis using AI
- Multimodal Machine Learning Applications
- Image Processing and 3D Reconstruction
- Face recognition and analysis
- Advanced Image Processing Techniques
- Topic Modeling
- Energy, Environment, and Transportation Policies
- Vehicle emissions and performance
Chongqing Jiulongpo People's Hospital
2025
Nanyang Technological University
2005-2024
State Key Laboratory of Information Engineering in Surveying Mapping and Remote Sensing
2024
Huazhong University of Science and Technology
2024
Nanjing University of Information Science and Technology
2024
Beijing Institute of Technology
2024
Chongqing Medical University
2024
Guizhou University
2023
University of British Columbia
2021-2023
Hunan University of Science and Technology
2023
Existing enhancement methods are empirically expected to help the high-level end computer vision task: however, that is observed not always be case in practice. We focus on object or face detection poor visibility enhancements caused by bad weathers (haze, rain) and low light conditions. To provide a more thorough examination fair comparison, we introduce three benchmark sets collected real-world hazy, rainy, low-light conditions, respectively, with annotated objects/faces. launched UG <sup...
Skeleton-based human action recognition has attracted increasing attention in recent years. However, most of the existing works focus on supervised learning which requiring a large number annotated sequences that are often expensive to collect. We investigate unsupervised representation for skeleton recognition, and design novel cloud colorization technique is capable representations from unlabeled sequence data. Specifically, we represent as 3D colorize each point according its temporal...
Abstract Non‐suicide self‐injury (NSSI) can be dangerous and difficult for guardians or caregivers to detect in time. NSSI refers when people hurt themselves even though they have no wish cause critical long‐lasting hurt. To timely identify effectively prevent order reduce the suicide rates of patients with a potential risk, detection based on spatiotemporal features indoor activities is proposed. Firstly, an behaviour dataset provided, it includes four categories that used scientific...
Few-shot classification aims to learn a discriminative feature representation recognize unseen classes with few labeled support samples. While most few-shot learning methods focus on exploiting the spatial information of image samples, frequency has also been proven essential in tasks. In this paper, we investigate effect different components To enhance performance and generalizability methods, propose novel Frequency-Guided Learning framework (dubbed FGFL), which leverages task-specific...
Recently, realistic DeepFake videos have raised severe security concerns in society. Existing video-based detection methods observe local spatial regions with the coarse temporal view, thus it is difficult to obtain subtle spatiotemporal information, resulting limited generalization ability. In this paper, we propose a novel Augmented Multi-scale Spatiotemporal Inconsistency Magnifier (AMSIM) Global View (GIV) and more meticulous Multi-timescale Local (MLIV), focusing on mining comprehensive...
Recently, DeepFake videos have developed rapidly, causing new security issues in society. Due to the rough spatiotemporal view, existing video-based detection methods struggle capture fine-grained information, resulting limited generalization ability. In addition, although transformer has achieved great success past few years, application of on deepfake video still needs be studied. To solve this problem, paper, we propose a novel Multiple Spatiotemporal Views Transformer (MSVT) with Local...
Deepfake techniques can forge the visual or audio signals in video, which leads to inconsistencies between and (VA) signals. Therefore, multimodal detection methods expose deepfake videos by extracting VA inconsistencies. Recently, technology has started collaborative forgery obtain more realistic videos, poses new challenges for Recent propose first extract natural correspondences real a self-supervised manner, then use learned as targets guide extraction of subsequent stage. However,...
3D Skeleton-based human action recognition has attracted increasing attention in recent years. Most of the existing work focuses on supervised learning which requires a large number labeled sequences that are often expensive and time-consuming to annotate. In this paper, we address self-supervised representation for skeleton-based recognition. We investigate design novel skeleton cloud colorization technique is capable spatial temporal representations from unlabeled sequence data. represent...
One-shot skeleton action recognition, which aims to learn a recognition model with single training sample, has attracted increasing interest due the challenge of collecting and annotating large-scale data. However, most existing studies match sequences by comparing their feature vectors directly neglects spatial structures temporal orders This paper presents novel one-shot technique that handles via multi-scale spatial-temporal matching. We represent data at multiple scales achieve optimal...
This paper reviews the second NTIRE challenge on image dehazing (restoration of rich details in hazy image) with focus proposed solutions and results. The training data consists from 55 images (with dense haze generated an indoor or outdoor environment) their corresponding ground truth (haze-free) same scene. has been produced using a professional haze/fog generator that imitates real conditions scenes. evaluation comparison dehazed images. process was learnable through provided pairs...
This study explores the dynamic deflection and vibrational analysis of a spherical shell, modeled as football game ball, reinforced with graphene platelet nanocomposites (GPLs). The leverages Carrera Unified Formulation (CUF) for accurate efficient modeling mechanical behavior shell under loads. CUF's flexibility in adapting to complex geometries material properties is utilized represent heterogeneous reinforcement GPLs within structure. To enhance reliability computational results, hybrid...
Few-Shot Class-Incremental Learning (FSCIL) aims to continuously learn new classes from a limited set of training samples without forgetting knowledge previously learned classes. Conventional FSCIL methods typically build robust feature extractor during the base session with abundant and subsequently freeze this extractor, only fine-tuning classifier in subsequent incremental phases. However, current strategies primarily focus on preventing catastrophic forgetting, considering relationship...
As one of the most important elements in intelligent transportation system (ITS), road traffic monitoring (RTMS) needs to be functioned with a recognition mechanism. Current works on mainly target at field automatic driving and cannot directly used RTMS. In this paper, we propose decision tree-based algorithm using roadside fixed light detection ranging (LiDAR) sensors These LiDAR have low vertical resolution, which implies that get clear far boundary obvious features roads from point cloud...
Traffic information collection is an important foundation for intelligent transportation systems. In this paper, 3D Light Detection And Ranging (LiDAR) deployed in the roadside of urban environments to collect vehicle and pedestrian information. A background filtering algorithm, including a mean modeling build map difference method filter static noise points, proposed fixed LiDAR facilities. Background points are filtered through between data frames multi-level map, then there still small...
Climate change and air pollution are two major environmental issues linked in several ways. Air pollutants reduction would benefit from the Greenhouse gas (GHG) emissions mitigation policy. China is facing a serious GHG emissions. Effective low cost strategies to solve these problems have been discussed by number of researchers. Previous studies typically evaluated near-term direct co-benefits, neglecting socio-economic impacts on CO2 PM2.5 emission. In this paper, production-based...
The state-of-the-art deep neural networks are vulnerable to common corruptions (e.g., input data degradations, distortions, and disturbances caused by weather changes, system error, processing). While much progress has been made in analyzing improving the robustness of models image understanding, video understanding is largely unexplored. In this paper, we establish a corruption benchmark, Mini Kinetics-C SSV2-C, which considers temporal beyond spatial images. We make first attempt conduct...