A5100419434- Advanced Image and Video Retrieval Techniques
- Human Pose and Action Recognition
- Face recognition and analysis
- Anomaly Detection Techniques and Applications
- Multimodal Machine Learning Applications
- Network Security and Intrusion Detection
- Advanced Neural Network Applications
- Face and Expression Recognition
- Generative Adversarial Networks and Image Synthesis
- Machine Learning and Data Classification
- Advanced Graph Neural Networks
- Machine Learning and Algorithms
- Internet Traffic Analysis and Secure E-voting
- Complex Network Analysis Techniques
- Text and Document Classification Technologies
- Advanced Malware Detection Techniques
- Video Surveillance and Tracking Methods
- Biometric Identification and Security
- Remote-Sensing Image Classification
- Adversarial Robustness in Machine Learning
- Topic Modeling
- Domain Adaptation and Few-Shot Learning
- Gait Recognition and Analysis
- Music and Audio Processing
- Advanced Steganography and Watermarking Techniques
Institute of Information Engineering
2016-2025
Chinese Academy of Sciences
2016-2025
Tianjin University
2023-2025
University of Chinese Academy of Sciences
2007-2024
Beijing Union University
2018-2024
Space Engineering University
2024
University of the Cordilleras
2024
Yanbian University
2024
Hebei Medical University
2024
Fourth Hospital of Hebei Medical University
2024
DeepFake detection aims to differentiate falsified faces from real ones. Most approaches formulate it as a binary classification problem by solely mining the local artifacts and inconsistencies of face forgery, which neglect relation across regions. Although several recent works explore learning for detection, they overlook propagation relational information lead limited performance gains. To address these issues, this paper provides new perspective formulating graph problem, in each facial...
This article presents FRCSyn-onGoing, an ongoing challenge for face recognition where researchers can easily benchmark their systems against the state of art in open common platform using large-scale public databases and standard experimental protocols. FRCSyn-onGoing is based on Face Recognition Challenge Era Synthetic Data (FRCSyn) organized at WACV 2024. first international aiming to explore use real synthetic data independently, also fusion, order address existing limitations technology....
Remote sensing image scene classification (RSI-SC) is crucial for various high-level applications, including RSI retrieval, captioning, and object detection. Deep learning-based methods can accurately predict categories. However, these approaches often require numerous labeled samples training, limiting their practicality in real-world RS applications with scarce label resources. In contrast, few-shot remote (FS-RSI-SC) has garnered substantial research interest owing to its potential...
Factorization machine (FM) is a prevalent approach to modelling pairwise (second-order) feature interactions when dealing with high-dimensional sparse data. However, on the one hand, FMs fail capture higher-order suffering from combinatorial expansion. On other taking into account between every pair of features may introduce noise and degrade prediction accuracy. To solve these problems, we propose novel approach, graph factorization (GraphFM), which naturally represents in structure. In...
In practical machine learning applications, human instruction is indispensable for model construction. To utilize the precious labeling effort effectively, active queries user with selective sampling in an interactive way. Traditional techniques merely focus on unlabeled data set under a unidirectional exploration framework and suffer from deterioration presence of noise. address this problem, paper proposes novel bidirectional algorithm that explores into both labeled sets simultaneously...
Action recognition in videos has attracted a lot of attention the past decade. In order to learn robust models, previous methods usually assume are trimmed as short sequences and require ground-truth annotations each video frame/sequence, which is quite costly time-consuming. this paper, given only video-level annotations, we propose novel weakly supervised framework simultaneously locate action frames well recognize actions untrimmed videos. Our proposed consists two major components....
Fast and accurate remote sensing image retrieval from large data archives has been an important research topic in the literature. Recently, hashing-based attracted extreme attention because of its efficient search capabilities. Especially, deep hashing algorithms have developed based on convolutional neural networks (CNNs) shown effective performance. However, implementing a network tends to be highly expensive terms storage space computing resources suitable for on-orbit retrieval, which...
Deep learning (DL) has been applied widely, and the quality of DL system becomes crucial, especially for safety-critical applications. Existing work mainly focuses on analysis models, but lacks attention to underlying frameworks which all models depend. In this work, we propose Audee, a novel approach testing localizing bugs. Audee adopts search-based implements three different mutation strategies generate diverse test cases by exploring combinations model structures, parameters, weights...
As a key marketing tool, online sales promotion has been widely used by retailers to increase of products and brands. Most previous researches on have ignored the effect consumers' psychological factors heterogeneity product consumers. The purpose this study is examine role distance involvement buying behavior in large activities. research model was examined using empirical analysis data obtained from consumer surveys after Double 11 promotion. Our results indicate that temporal positive...
Substantial improvements have been achieved in the field of face recognition due to successful application deep neural networks. However, existing methods are sensitive both quality and quantity training data. Despite availability large-scale datasets, long tail data distribution induces strong biases model learning. In this paper, we present a Face Augmentation Generative Adversarial Network (FA-GAN) reduce influence imbalanced deformation attribute distributions. We propose decouple these...
Abstract The rising demand of strategic metals, especially lithium, necessitates discovery new resources to meet the global supply chain. Recently, several pegmatite-hosted rare metal (Li-Rb-Be-Nb-Ta) deposits have been discovered in Western Kunlun orogenic belt, making it a world-class resource (estimated ~7 Mt Li2O and 0.16 BeO). Understanding metallogenesis this belt is critical further evaluate potential. In study, columbite-tantalite (coltan) monazite from pegmatites zircon potential...
Detecting digital face manipulation has attracted extensive attention due to fake media's potential risks the public. However, recent advances have been able reduce forgery signals a low magnitude. Decomposition, which reversibly decomposes an image into several constituent elements, is promising way highlight hidden details. In this paper, we investigate novel 3D decomposition based method that considers as production of interaction between geometry and lighting environment. Specifically,...
Click-through rate (CTR) prediction is an essential task in web applications such as online advertising and recommender systems, whose features are usually multi-field form. The key of this to model feature interactions among different fields. Recently proposed deep learning based models follow a general paradigm: raw sparse input first mapped into dense field embedding vectors, then simply concatenated together feed neural networks (DNN) or other specifically designed learn high-order...
The point process is a solid framework to model sequential data, such as videos, by exploring the underlying relevance. As challenging problem for high-level video understanding, weakly supervised action recognition and localization in untrimmed videos have attracted intensive research attention. Knowledge transfer leveraging publicly available trimmed external guidance promising attempt make up coarse-grained video-level annotation improve generalization performance. However, unconstrained...
With machine learning models especially Deep Neural Network (DNN) becoming an integral part of the new intelligent software, tools to support their engineering process are in high demand. Existing DNN debugging either post-training which wastes a lot time training buggy model and requires expertises, or limited on collecting logs without analyzing problem not even fixing them. In this paper, we propose AUTOTRAINER, monitoring automatic repairing tool supports detecting auto five commonly...
Graph embedding, aiming to learn low-dimensional representations (aka. embeddings) of nodes in graphs, has received significant attention. In recent years, there been a surge efforts, among which graph convolutional networks (GCNs) have emerged as an effective class models. However, these methods mainly focus on the static embedding. present work, efficient dynamic embedding approach is proposed, called GCN (DyGCN), extension GCN-based methods. The propagation scheme naturally generalized...
Inonotus obliquus is a medicinal fungus that has been favored for its anti-diabetic benefits in China thousands of years. Inotodiol characteristic triterpenoid isolated from obliquus. The inactivation effect and mechanism α-glucosidase induced by inotodiol were investigated using fluorescence, FTIR, CD spectroscopy, molecular dynamics simulation. could effectively inhibit activity (IC50 = 1.09 ± 0.08 mmol/L) reversible noncompetitive manner. interacted with via static quenching process...
Download This Paper Open PDF in Browser Add to My Library Share: Permalink Using these links will ensure access this page indefinitely Copy URL DOI
Contemporary deep face recognition techniques predominantly utilize the Softmax loss function, designed based on similarities between sample features and class prototypes. These can be categorized into four types: in-sample target similarity, non-target out-sample similarity. When a feature from specific is designated as anchor, similarity this any prototype referred to In contrast, samples other classes known The terms indicate whether used for calculation belong same identity or not....