A5100437641- Advanced Neural Network Applications
- Video Surveillance and Tracking Methods
- Visual Attention and Saliency Detection
- AI in cancer detection
- Graph Theory and Algorithms
- Generative Adversarial Networks and Image Synthesis
- Advanced Vision and Imaging
- Image and Video Quality Assessment
- Infrared Target Detection Methodologies
- Traffic Prediction and Management Techniques
- Medical Image Segmentation Techniques
- Simulation and Modeling Applications
- Transportation Planning and Optimization
- Vehicle License Plate Recognition
- Image Enhancement Techniques
- Video Coding and Compression Technologies
- Safety Warnings and Signage
- Spectroscopy Techniques in Biomedical and Chemical Research
- Machine Learning and Data Classification
- Metallic Glasses and Amorphous Alloys
- CCD and CMOS Imaging Sensors
- Currency Recognition and Detection
- Underwater Acoustics Research
- Advanced Measurement and Detection Methods
- Soil Moisture and Remote Sensing
Shenyang University of Technology
2024
Yangzhou University
2024
Zhejiang University of Science and Technology
2023
Macau University of Science and Technology
2021-2023
Great Wall Motors (China)
2023
Chinese Academy of Sciences
2022
University of Chinese Academy of Sciences
2022
China Agricultural University
2022
Tiangong University
2022
Ningbo Institute of Industrial Technology
2022
SAR ship detection is sensitive to the interference of inshore background, disturbance strong wind and waves. The similar textures neighbor objects in images affect performance. As a remarkable indicator, textural information frequency domain characterizes subtle differences between an object its surroundings. Inspired by this, multi-feature fusion network (MFFN) for constructed this paper, which can obtain contour detail image detecting ships from their background. Firstly, spatial targets,...
Periodic graphs are consisting of repetitive local structures, such as crystal nets and polygon mesh. Their generative modeling has great potential in real-world applications material design graphics synthesis. Classical models either rely on domain-specific predefined generation principles (e.g., net design), or follow geometry-based prescribed rules. Recently, deep shown promise automatically generating general graphs. However, their advancement into periodic have not been well explored...
Learning a second language is challenging. Becoming fluent requires learning contextual information about how should be used as well word meanings and grammar. The majority of existing applications provide only thin context around content. In this paper, we present work in Crystallize, game that combines traditional approaches with situated paradigm by integrating spaced-repetition system within roleplaying game. To facilitate long-term engagement the game, added new quest paradigm, "jobs,"...
Spaceborne infrared ship detection holds immense research significance in both military and civilian domains. Nonetheless, the focus of this field remains primarily on optical synthetic aperture radar (SAR) images due to confidentiality limited accessibility data. The challenges spaceborne arise from long-distance capture low signal-to-noise ratio images, which contribute false alarm misclassifications. To handle problem, article concentrates enhancing information interaction during feature...
In order to improve the computational efficiency of convolutional neural networks (CNNs) for object detection on reconfigurable platforms such as field-programmable gate arrays (FPGAs), we propose a CNN processor with hierarchical pipelining and multicore computing based parallel parameter constraints. First, pipelined processing architecture that can adapt computations on-chip memory requirements different layers. We present design configure units while adjusting interconnection utilization...
Developing deep generative models has been an emerging field due to the ability model and generate complex data for various purposes, such as image synthesis molecular design. However, advancement of is limited by challenges objects that possess multiple desired properties: 1) existence correlation among real-world properties common but hard identify; 2) controlling individual property enforces implicit partially control its correlated properties, which difficult model; 3) under manners...
The traditional tourism system’s information inquiry is incomplete, images are not intuitive and the results three-dimensional. In order to solve these problems, this paper proposes system management design application based on tourist routes inquiry. Firstly, MAP technology, it achieves a cross-check of space attribute data, then explores knapsack problem through genetic algorithm, making travel route arrangements more reasonable providing latest shareable maps for tourists, finally...
In aerial photography, UAV will cause small objects and mutual occlusion between objects, bringing enormous challenges to the current one-stage object detection algorithm. order solve above problems, this paper proposes a YOLOX-TBP Based on YOLOX-l, algorithm adds prediction head detect of different scales improve effect objects. At same time, partial convolution backbone network is replaced by Transformer global information model, weighted bi-directional feature pyramid (BiFPN) introduced...
Semi-supervised instance segmentation algorithms are mainly divided into based on pseudo-label generation and transfer learning. The need to design a specific process, but the process is not scalable for different types of source tasks. learning that started late have relatively high scalability, algorithm research ideas simple. To expand semi-supervised learning, this paper proposes feature transfer-based Feature Transfer Mask R-CNN (FT-Mask). FT-Mask more than can be used knowledge from...
In order to maintain the overall traffic orders of urban transport system, prevent network jam and second accident occur during roadway emergency incidents, control mechanisms at scenes are set up. Firstly, basing on analysis spots in dynamic environment, zone constructed, which include early-warning area, approach transition activity area termination area. Secondly, by means selection vehicles interception positions, length taper lines zone, mechanism formed. Thirdly, setting demands...
Traffic flow prediction, as an integral part of intelligent transportation systems, is carried out by using data from the traffic sensors in network. In a typical prediction for specific location, most essential has traditionally been recent records obtained that same location. However, are prone to failure, and result measurements location interest may be unavailable. this work, we propose model predicts locations with faulty sensors, solely neighboring sensors. We use online recursive...
The recycling of discarded bottles can save resources and avoid pollution. current machinery is faced with high cost or large manual operation. Furthermore, it cannot deal complex classification types environmental problems. A designment sorting system couple Res2net combining attention mechanism label smoothing methods identified classified for different post-consumer bottles, the result shows that recognition accuracy this way has been improved. method not only reduce data disturbance at...
Writer identification has practical applications for forgery detection and forensic science. Most models based on deep neural networks extract features from character image or sub-regions in image, which ignoring contained page-region image. Our proposed global regular network (GRN) pays attention to these features. GRN consists of two branches: one branch takes page handwriting as input features, the other word local Global merge a residual way form overall handwriting. The attributions: is...
Dataset is important in facial expression recognition, but they may run into apparent imbalances caused by various collection scenes. This paper studies the class imbalance problem benchmark dataset for recognition. We focus on design of focal loss function instead cross entropy to solve imbalance. Graph convolutional network (GCN) applied achieve Experimental results Oulu-CASIA and SFEW 2.0 exhibit that proposed programme improves performances
With the full coverage of highway network in each country, workload construction and maintenance is getting bigger bigger. The establishment work zones has brought a more complex road environment for motorists. A large number injuries fatalities occur zones, which about 60% are estimated to be drivers. Many accident analysis studies have also shown that these accidents mainly caused by driver's own human factors inadequate safety facilities. In order reduce occurrence accidents, some...