A5100637864- Advanced Vision and Imaging
- Human Pose and Action Recognition
- Advanced Neural Network Applications
- Microbial Metabolites in Food Biotechnology
- Gait Recognition and Analysis
- Privacy-Preserving Technologies in Data
- Video Surveillance and Tracking Methods
- Image Processing Techniques and Applications
- Context-Aware Activity Recognition Systems
- Enzyme Production and Characterization
- Topic Modeling
- Multimodal Machine Learning Applications
- Advanced Image Processing Techniques
- Diet, Metabolism, and Disease
- Advanced Text Analysis Techniques
- Optical measurement and interference techniques
- IoT and Edge/Fog Computing
- Educational Technology and Pedagogy
- Hand Gesture Recognition Systems
- Cognitive Science and Mapping
- Molecular Communication and Nanonetworks
- Plant nutrient uptake and metabolism
- Age of Information Optimization
- Robotics and Automated Systems
- Advanced Algorithms and Applications
Chongqing University
2025
Jiangsu University
2024
Beijing Jiaotong University
2022-2024
Xidian University
2023-2024
University of St. Gallen
2023
Jiangnan University
2018-2022
State Key Laboratory of Food Science and Technology
2018-2022
Beijing Normal University - Hong Kong Baptist University United International College
2022
Southeast University
2022
China University of Petroleum, East China
2021
GCN-based methods have achieved remarkable performance in skeleton-based action recognition. However, existing not explicitly attempted to remove temporal and spatial redundancy that might introduce additional computational costs. Inspired by the fact humans always tend glimpse at overall motion then zoom into most important spatio-temporal regions, we propose a Spatio Temporal Focused Dynamic Network (STFD-Net) trained with reinforcement learning for Specifically, first global extractor...
Recent work has highlighted the risks of LLM-generated content for a wide range harmful behaviors, including incorrect and code. In this work, we extend by studying whether web design contains dark patterns. This evaluated designs ecommerce components generated four popular LLMs: Claude, GPT, Gemini, Llama. We tested 13 commonly used (e.g., search, product reviews) them as prompts to generate total 312 across all models. Over one-third contain at least one pattern. The majority pattern...
This paper presents a comprehensive underwater visual reconstruction paradigm that comprises three procedures, i.e., the E-procedure, R-procedure, and H-procedure. The E-procedure enhances original images based on color compensation balance weighted image fusion, yielding restored color, sharpened edges, global contrast. R-procedure registers multiple enhanced by exploiting similarity local deformation. H-procedure homogenizes registered multi-scale composition strategy, which eliminates...
Self-supervised monocular methods can efficiently learn depth information of weakly textured surfaces or reflective objects. However, the accuracy is limited due to inherent ambiguity in geometric modeling. In contrast, multi-frame estimation improve thanks success Multi-View Stereo (MVS), which directly makes use constraints. Unfortunately, MVS often suffers from texture-less regions, non-Lambertian surfaces, and moving objects, especially real-world video sequences without known camera...
In recent years, vision-centric perception has flourished in various autonomous driving tasks, including 3D detection, semantic map construction, motion forecasting, and depth estimation. Nevertheless, the latency of approaches is too high for practical deployment (e.g., most camera-based detectors have a runtime greater than 300ms). To bridge gap between ideal researches real-world applications, it necessary to quantify trade-off performance efficiency. Traditionally, autonomous-driving...
Pedestrian detection provides manager of a smart city with great opportunity to manage their effectively and automatically. Specifically, pedestrian technology can improve our secure environment make traffic more efficient. In this paper, all work both modification improvement are made based on YOLO, which is real-time Convolutional Neural Network detector. work, we extend YOLO’s original network structure, also give new definition loss function boost the performance for detection,...
Ship detection is of great value for fishing activity control, military defense, maritime transport, etc. Satellite-based synthetic aperture radar (SAR) can provide high-resolution images, allowing surveillance over massive water bodies to be possible. However, traditional ship algorithms like CFAR (Constant False-Alarm Rate), cannot produce convincing results. In recent years, detectors based on convolutional neural networks have made progress, and among them, Faster R-CNN one the best in...
In this research project, we used the financial texts published by Federal Open Market Committee (FOMC), known as FOMC Minutes, for sentiment analysis. The pre-trained FinBERT model, a state-of-the-art transformer-based model trained NLP tasks in finance, was utilized that. focus of has been on improving predictive performance complex sentences, our problem analysis shown that such sentences pose significant challenge to existing models. To accomplish objective original fine-tuned...
Federated machine learning is a versatile and flexible tool to utilize distributed data from different sources, especially when communication technology develops rapidly an unprecedented amount of could be collected on mobile devices nowadays. method exploits not only the but computational power all in network achieve more efficient model training. Nevertheless, while most traditional federated methods work well for homogeneous tasks, adapting heterogeneous task distribution challenging....
Falls constitute a significant health risk, particularly among the elderly, thus prompting introduction of various wearable devices capable fall detection. However, majority these prioritize accuracy over wearer comfort, which significantly influences user adherence and, by extension, broader development technologies. Addressing this oversight, review first summarizes current methods for predicting comfort devices, evaluating them in terms feasibility and accuracy, reliability effectiveness,...
Realizing scaling laws in embodied AI has become a focus. However, previous work been scattered across diverse simulation platforms, with assets and models lacking unified interfaces, which led to inefficiencies research. To address this, we introduce InfiniteWorld, scalable simulator for general vision-language robot interaction built on Nvidia Isaac Sim. InfiniteWorld encompasses comprehensive set of physics asset construction methods generalized free benchmarks. Specifically, first...
Parallel text datasets are a valuable for educational purposes, machine translation, and cross-language information retrieval, but few domain-oriented. We have created Chinese–English parallel dataset in the domain of finance technology, using <em>Financial Times</em> website, from which we grabbed 60,473 news items between 2007 2021. This is bilingual finance. It open access its original state without transformation, has been made not translation as used, intelligent mining, conducted many...