- Human Pose and Action Recognition
- Smart Agriculture and AI
- Visual Attention and Saliency Detection
- Hand Gesture Recognition Systems
- Robot Manipulation and Learning
- Generative Adversarial Networks and Image Synthesis
- Face recognition and analysis
- Remote Sensing and Land Use
- Plant Disease Management Techniques
- Advanced MRI Techniques and Applications
- Advanced Neural Network Applications
- Infrared Target Detection Methodologies
- Olfactory and Sensory Function Studies
- IoT and Edge/Fog Computing
- Fire Detection and Safety Systems
- Grief, Bereavement, and Mental Health
- Remote Sensing in Agriculture
- Spectroscopy and Chemometric Analyses
- Shoulder Injury and Treatment
- Horticultural and Viticultural Research
- Functional Brain Connectivity Studies
- Anatomy and Medical Technology
- Knee injuries and reconstruction techniques
- Retinal Imaging and Analysis
- Image and Object Detection Techniques
University of South China
2025
Jiangxi Agricultural University
2009-2024
Xi'an Jiaotong University
2024
Zhejiang University
2020-2021
Yanshan University
2014-2015
National Cheng Kung University
2011
With the development of deep learning, synthetic aperture radar (SAR) image ship detection based on convolutional neural network has made significant progress. However, there are two problems. 1) The false alarm rate is high due to complex background and coherent speckle noise interference. 2) For smaller targets, missed prone occur. In this letter, a novel model (MFTF-Net) multi-feature transformation fusion proposed address issues. First, avoid randomness initial point selection influence...
As one of the most important economic crops, grapes have attracted considerable attention due to their high yield, rich nutritional value, and various health benefits. Identifying grape bunches is crucial for maintaining quality quantity grapes, as well managing pests diseases. In recent years, combination automated equipment with object detection technology has been instrumental in achieving this. However, existing lightweight algorithms often sacrifice precision processing speed, which may...
Background: The abnormal morphology of complete discoid lateral meniscus (CDLM) is associated with a greater propensity for tears. magnetic resonance imaging (MRI) CDLM most often described in the coronal plane, few morphological studies sagittal position. Hypothesis: anteroposterior diameter smaller than that normal and increased after tear adults. Study Design: Cross-sectional study; Level evidence, 3. Methods: authors searched radiology records at our institution from June 2018 to...
Blueberries have high nutritional and economic value are easy to cultivate, so they common fruit crops in China. There is a demand for blueberry domestic foreign markets, various technologies been used extend the supply cycle of about 7 months. However, grows clusters, cluster fruits generally contains different degrees maturity, which leads low efficiency manually picking mature fruits, at same time wastes lot manpower material resources. Therefore, order improve efficiency, it necessary...
Blueberries are widely planted because of their rich nutritional value. Due to the problems dense adhesion and serious occlusion blueberries during growth process, development automatic blueberry picking has been seriously hindered. Therefore, using deep learning technology achieve rapid accurate positioning in case is one key technologies blueberries. To improve accuracy, this paper designs a recognition model based on improved YOLOv5. Firstly, dataset constructed. On basis, we design new...
The damage caused by pests to crops results in reduced crop yield and compromised quality. Accurate timely pest detection plays a crucial role helping farmers defend against control pests. In this paper, novel model named YOLOv5s-pest is proposed. Firstly, we design hybrid spatial pyramid pooling fast (HSPPF) module, which enhances the model’s capability capture multi-scale receptive field information. Secondly, new convolutional block attention module (NCBAM) that highlights key features,...
Estimating accurate 3D hand pose from a single RGB image is highly challenging problem in estimation due to self-geometric ambiguities, self-occlusions, and the absence of depth information. To this end, novel Five-Layer Ensemble CNN (5LENet) proposed based on hierarchical thinking, which designed decompose task into five single-finger sub-tasks. Then, sub-task results are fused estimate full pose. The method great benefit extract deeper better finger feature information, can effectively...
Salient region detection is important for many computer vision tasks. The saliency results may serve as the basis further high‐level tasks like object segmentation and tracking. In this study, authors propose an integration approach to detect salient based on three principles from psychological evidence observations of images, including colour contrast in a global context, spatially compact distribution, multi‐scale image abstraction. Based above‐mentioned principles, authors’ analysis can...
3D hand pose estimation can provide basic information about gestures, which has an important significance in the fields of Human-Machine Interaction (HMI) and Virtual Reality (VR). In recent years, from a single depth image made great research achievements due to development cameras. However, RGB is still highly challenging problem. this work, we propose novel four-stage cascaded hierarchical CNN (4CHNet), leverages network decompose into finger palm estimation, extracts separately features...
Currently, deep convolutional neural networks have achieved great achievements in semantic segmentation tasks, but existing methods all require a large number of annotated images for training and do not good scalability new objects. Therefore, few-shot that can identify objects with only one or few are gradually gaining attention. However, the current cannot segment plant diseases well. Based on this situation, disease model multi-scale multi-prototypes match (MPM) is proposed. This method...
Most existing methods reconstruct 3D hand mesh based on artificial prior knowledge, which exist great contingency and aren’t robust. To this end, we propose GPFormer, can simultaneously capture apparent topologies, implicit semantics, the global local interactions of in an autonomous manner. Based it, a novel two-stage pipeline is constructed, including Pose GPFormer Mesh PFormer. Specifically, consists multiple GPFormers with fixed learnable graph convolution filters locating parallel...
We introduce a novel approach to detect salient regions of an image via feature combination and discriminative classifier. Our method, which is based on hierarchical abstraction, uses the logistic regression map regional vector saliency score. Four cues are used in our approach, including color contrast global context, center-boundary priors, spatially compact distribution, objectness, as atomic segmented region image. By mapping four-dimensional fifteen-dimensional vector, we can linearly...
Due to the sophisticated entanglements for non-rigid deformation, generating person images from source pose target is a challenging work. In this paper, we present novel framework generate with shape consistency and appearance consistency. The proposed leverages graph network infer global relationship of in better transfer. Moreover, decompose image into different attributes (e.g., hair, clothes, pants shoes) combine them coding through operation method more realistic image. We adopt an...
Arterial spin labeling (ASL) is a non-invasive magnetic resonance imaging (MRI) method that can provide direct and quantitative measurements of cerebral blood flow (CBF) scanned patients. ASL be utilized as an modality to detect Alzheimer's disease (AD), brain atrophy AD patients revealed by low CBF values in certain regions. However, partial volume effects (PVE), which mainly caused signal cross-contamination due voxel heterogeneity limited spatial resolution images, often prevents from...
Advance Care Planning (ACP) is a national social movement. The purpose of this study to evaluate the effectiveness life educational program and make recommendations related its delivery. subjects were 252 students between ages 7 13 from one elementary school in southern Taiwan. objective education introduce life-cycle concepts that would help student learn coping skills for painful events. lasted 40 min was conducted as follows: (1) A brief introduction rapport building (10 min); (2)...
In order to construct a color model of rice leaf based on physiology and ecology, modeling method SVM BP neural network was proposed for the relationship between chlorophyll, carotenoid its RGB value. The chlorophyll a, b were used as input parameters, R, G B values image output parameters respectively corresponding component predicted by using network. results show that prediction accuracy is significantly higher than SVM. research can meet needs agriculture provide theoretical basis...
A machine learning method was used to establish a diagnostic model of potassium nutrition for rice obtained from image processing techniques.In this study, super hybrid "Liangyoupeijiu" as experimental object set up four kinds cultivation experiments at different fertilization levels, the data total 1920 groups 1st leaves and 2nd leaves, 3rd their corresponding sheaths were by scanning with scanner.Nineteen characteristic indexes obtained.Support vector in nineteen indexes, diagnose identify...