A5100774574- Higher Education and Teaching Methods
- Multimodal Machine Learning Applications
- Advanced Computational Techniques and Applications
- Human Pose and Action Recognition
- Simulation and Modeling Applications
- Education and Work Dynamics
- Educational Technology and Assessment
- Natural Language Processing Techniques
- Anomaly Detection Techniques and Applications
- Chinese history and philosophy
- Geomechanics and Mining Engineering
- Civil and Geotechnical Engineering Research
- Domain Adaptation and Few-Shot Learning
- Advanced Image and Video Retrieval Techniques
- Traditional Chinese Medicine Analysis
- Evaluation and Optimization Models
- Topic Modeling
- Cosmology and Gravitation Theories
- Remote Sensing and Land Use
- Dark Matter and Cosmic Phenomena
- Experimental Learning in Engineering
- Particle physics theoretical and experimental studies
- Speech and Audio Processing
- Music and Audio Processing
- Advanced Neural Network Applications
Guangxi Normal University
2002-2025
Shanghai Jiao Tong University
2020-2025
Alibaba Group (China)
2024-2025
Guizhou Academy of Agricultural Sciences
2025
Kai Biotech (South Korea)
2025
China Agricultural University
2025
Chengdu University of Traditional Chinese Medicine
2017-2024
Shandong Jiaotong University
2020-2024
Tsinghua–Berkeley Shenzhen Institute
2024
Tsinghua University
2015-2024
The objective of Audio-Visual Segmentation (AVS) is to localise the sounding objects within visual scenes by accurately predicting pixel-wise segmentation masks. To tackle task, it involves a comprehensive consideration both data and model aspects. In this paper, first, we initiate novel pipeline for generating artificial AVS task without extra manual annotations. We leverage existing image audio datasets match image-mask pairs with its corresponding samples using category labels in...
This study is aimed at exploring whether learning about art activities can help children of divorced parents in three aspects, namely self-concept, emotional expression, and linguistic expression.The researcher designed appropriate for based on their family backgrounds.There were nine total, including "self the mirror", "painting with paint", "pick-up paste".From February to April 2016, these conducted once a week 40 minutes, led by researcher.During activities, data collected through...
One of the main problems in representing information form nonsystematic logic is lack flexibility, which leads to potential overfitting. Although improves representation conventional k Satisfiability, formulations first, second, and third-order logical structures are very predictable. This paper proposed a novel higher-order structure, named G-Type Random by capitalizing new random feature clauses. The was implemented into Discrete Hopfield Neural Network as symbolic rule. Networks evaluated...
This work presents ChainFS, a middleware system that secures cloud storage services using minimally trusted Blockchain. ChainFS hardens the cloud-storage security against forking attacks. The exposes file-system interface to end users. Internally, stores data files in and exports minimal necessary functionalities Blockchain for key distribution file operation logging. We implement on Ethereum S3FS closely integrate it with FUSE clients Amazon S3 storage. measure performance demonstrate low overhead.
Single-frame temporal action localization (STAL) aims to localize actions in untrimmed videos with only one timestamp annotation for each instance. Existing methods adopt the one-stage framework but couple counting goal and goal. This paper proposes a novel two-stage STAL task spirit of divide conquer. The instance stage leverages location supervision determine number instances whole video into multiple clips, so that clip contains complete instance; estimation category clip. To efficiently...
Weakly-supervised temporal action localization (WTAL) learns to detect and classify instances with only category labels. Most methods widely adopt the off-the-shelf Classification-Based Pre-training (CBP) generate video features for localization. However, different optimization objectives between classification localization, make temporally localized results suffer from serious incomplete issue. To tackle this issue without additional annotations, paper considers distill free knowledge...
Cross-domain CTR (CDCTR) prediction is an important research topic that studies how to leverage meaningful data from a related domain help in target domain. Most existing CDCTR works design implicit ways transfer knowledge across domains such as parameter-sharing regularizes the model training More effectively, recent researchers propose explicit techniques extract user interest and this However, proposed method mainly faces two issues: 1) it usually requires super domain, i.e. extremely...
Spin-2 ultralight dark matter (ULDM) is a viable candidate and it can be constrained using gravitational wave (GW) observations. In this paper, we investigate the detectability of spin-2 ULDM by space-based GW interferometers. By considering direct coupling between ordinary matter, derive corresponding response functions sensitivity curves for various time-delay interferometry channels calculate optimal future millihertz detectors. Our results demonstrate that detectors place stringent...
Temporal sentence grounding aims to detect event timestamps described by the natural language query from given untrimmed videos. The existing fully-supervised setting achieves great results but requires expensive annotation costs; while weakly-supervised adopts cheap labels performs poorly. To pursue high performance with less costs, this paper introduces an intermediate partially-supervised setting, i.e., only short-clip is available during training. make full use of partial labels, we...
The chiral gravitational wave background (GWB) can be produced by axion-like fields in the early universe. We perform parameter estimation for two types of GWB with LISA-Taiji network: axion-dark photon coupling and axion-Nieh-Yan coupling. estimate spectral parameters these mechanisms induced axion determine normalized model using Fisher information matrix. For highly signals that we choose to analyze mHz band, are constrained a relative error less than $6.7\%$ (dark coupling) $2.2\%$...
The Internet of Things (IoT) has been developing rapidly in the past few years. In IoT, an enormous number smart devices are connected to network, where communication and interaction occurs extensively among end users, services. Due great diversity devices, broader scope interactions other characteristics current IdM model for needs be extended improved. objective this article is analyze main features IoT key issues then present framework which consists three parts: standard information...
It has been shown that species diversity of tree frogs in the genus Zhangixalus might have underestimated. In this study, based on previously published data and newly collected samples from China, we constructed a phylogeny using mitochondrial 12S rRNA-tRNA Val -16S rRNA genes revealed two novel lineages southern one containing Pingbian, Yunnan, Chengbu, Hunan. Combining evidence morphological comparison delimitation, described these as new species. daweishanensis sp. nov. Mt. Dawei,...
In this paper, we investigate the generation of new video backgrounds given a human foreground video, camera pose, and reference scene image. This task presents three key challenges. First, generated background should precisely follow movements corresponding to foreground. Second, as shifts in different directions, newly revealed content appear seamless natural. Third, objects within frame maintain consistent textures moves ensure visual coherence. To address these challenges, propose...
Atherosclerosis (AS) is a major and common pathological basis of ischemic intestinal infarction, myocardial stroke, renal failure, other highly lethal disabling diseases. Current pharmacological interventions (e.g., statins) often cause adverse effects, limiting their long-term use. Natural compounds, with multitarget efficacy superior safety profiles, have emerged as promising alternatives for AS treatment. As potent antioxidant carotenoid, astaxanthin exhibits unique therapeutic potential...
Abstract The chiral gravitational wave background (GWB) can be produced by axion-like fields in the early universe. We perform parameter estimation for two types of GWB with LISA-Taiji network: axion-dark photon coupling and axion-Nieh-Yan coupling. estimate spectral parameters these mechanisms induced axion determine normalized model using Fisher information matrix. For highly signals that we choose to analyze mHz band, are constrained a relative error less than $6.7\%$ (dark coupling)...