A5044293662- Advanced Image and Video Retrieval Techniques
- Vehicle License Plate Recognition
- Handwritten Text Recognition Techniques
- Image and Signal Denoising Methods
- Remote-Sensing Image Classification
- Advanced Neural Network Applications
- Domain Adaptation and Few-Shot Learning
- Mobile Ad Hoc Networks
- Multi-Agent Systems and Negotiation
- Kawasaki Disease and Coronary Complications
- Non-Invasive Vital Sign Monitoring
- Video Surveillance and Tracking Methods
- Higher Education Learning Practices
- Digital Rights Management and Security
- Human Pose and Action Recognition
- Machine Learning and ELM
- Higher Education and Teaching Methods
- Advanced Decision-Making Techniques
- Image Processing and 3D Reconstruction
- Recommender Systems and Techniques
- Video Analysis and Summarization
- Model-Driven Software Engineering Techniques
- Education and Vocational Training
- Face and Expression Recognition
- Semantic Web and Ontologies
Jiaxing University
2007-2023
Shanghai University
2016-2018
Detecting irregular or arbitrary shape text in natural scene images is a challenging task that has recently attracted considerable attention from research communities. However, limited by the CNN receptive field, these methods cannot directly capture relations between distant component regions local convolutional operators. In this paper, we propose novel method can effectively and robustly detect images. First, employ fully network architecture based on VGG16_BN to generate components via...
Aerial scene classification is a challenging task in the remote sensing image processing field. Owing to some similar scene, there are only differences density. To challenge this problem, paper proposes novel parallel multi-stage (PMS) architecture formed by low, middle, and high deep convolutional neural network (DCNN) sub-model. PMS model automatically learns representative discriminative hierarchical features, which include three 512 dimension vectors, respectively, final feature created...
Identifying a discriminative feature can effectively improve the classification performance of aerial scene classification. Deep convolutional neural networks (DCNN) have been widely used in for its learning ability. The DCNN be more by optimizing training loss function and using transfer methods. To enhance power feature, improved functions pretraining models are combined with softmax centre function. further performance, this article, we propose hybrid features First, use joint from...
Recommendation systems are widely used to cope with the problem of information overload and, consequently, many recommendation methods have been developed for present systems, such as content-based, collaborative filtering, Web mining-based and so on. But they always lack intelligence, self-adaptiveness initiative. Aiming at these disadvantages, in this work, a personalized system (APRS) is presented multi-agents based on intelligence though. This paper discusses structure APRS first. In...
The bottleneck has a faster energy consumption rate than other nodes, and it may lead to network partition in wireless sensor networks. To save the of bottleneck, we suggested coarse-grained detection method based on data path for multi-hop Moreover, buffer mechanism around was introduced reduce control frames media access layer. Experiments showed our scheme can prolong lifetime, but also caused little longer delay.
Unsupervised domain adaptation uses source data from different distributions to solve the problem of classifying unlabeled target domains. However, conventional methods require access data, which often raise concerns about privacy. In this paper, we consider a more practical but challenging setting where is unavailable and unlabeled. Specifically, address discrepancy perspective contrastive learning. The key idea our work learn domain-invariant feature by 1) performing clustering directly in...
With the increasing application based on Web, comparatively backward Web printing technology lead so much inconvenience to development of system and slowed pace in enterprises. In this work, by making use component programming ideas EJB technology, a print components was developed which analysis PDF format. Thus, it provides simple practical solution component. The schema especially meets requirements with fixed format changing content.
The education of talents in software industry faces both opportunities and challenges, because the fast update for knowledge, extensive use fields technologies, constant emergence new technologies software. In this work , a model cultivating is presented after giving brief introduction about current situation market demand such talented men traditional educational pattern talents, characterized by teaching - project-oriented under tutor. And then describes it from seven aspects: goal,...