A5028173524- Advanced Neural Network Applications
- Advanced Image and Video Retrieval Techniques
- Remote-Sensing Image Classification
- Video Surveillance and Tracking Methods
- Retinal Imaging and Analysis
- Domain Adaptation and Few-Shot Learning
- Adversarial Robustness in Machine Learning
- Multimodal Machine Learning Applications
- Morphological variations and asymmetry
- Railway Engineering and Dynamics
- Simulation and Modeling Applications
- Gaze Tracking and Assistive Technology
- Brain Tumor Detection and Classification
- Image Retrieval and Classification Techniques
- Visual Attention and Saliency Detection
- Glaucoma and retinal disorders
- Advanced Measurement and Detection Methods
Guizhou University
2025
Southwest University of Science and Technology
2023-2024
Guangdong Ocean University
2024
Mianyang Normal University
2024
Ritsumeikan University
2015
Abstract Object detection plays a vital role in remote sensing applications. Although object has achieved proud results natural images, these methods are difficult to be directly applied images. Remote images often have complex backgrounds and small objects, which highly unbalanced distribution of foreground background information. In order solve the above problems, this paper proposes multi-head channel spatial trans-attention (MCSTA) module, performs pixel interaction from dimensions...
Vision-language pre-training models have achieved significant success in the field of medical imaging but exhibited vulnerability to adversarial examples. Although attacks are harmful, they valuable revealing weaknesses VLP and enhancing their robustness. However, due under-utilization modal differences consistent features existing methods, attack effectiveness migration samples not satisfactory. To address this issue enhance transferability, we propose multimodal feature heterogeneous...
Retinal diseases impair the normal function of visual system, making accurate segmentation retinal vessels crucial. This paper proposes an improved U-Net network, namely Mitigating Information Loss (MILU-Net), for vessel segmentation. The network introduces Multi-Scale Adaptive Detail Feature Fusion (MSADFF) module, ensuring effective fusion features at different scales before skip connections to reduce information loss. Simultaneously, Dual Path Upsampling (DPUS) module is employed enhance...
To at the low robustness of existing model for occluded object detectiont, an detection algorithm based on fuzzy sample anchor box IoU Matching degree Deviation Aware (IoU_MDA) is proposed. Firstly, samples are defined Anchor-based, which reflects occlusion. Secondly, IoU_MDA proposed to quantify interference experienced by samples. Then, IoU_MDA_Loss constructed IoU_MDA, combined with and balance parameterΦ. address class imbalance issues enhance generality, intra-class inter-class weights,...
With the increase in port throughput and development of trend large-scale ships, selecting applicable anchor positions for ships ensuring rational comprehensive utilization anchorage areas have become a key issue utilizing rate resources, safety anchoring operations promoting shipping industry. Existing position selection detection algorithm studies are limited to two-dimensional plane ship selection, with few considering intelligent algorithms safe water depths based on three-dimensional...
Recently, the genetic association of human facial morphological variation attracts substantial attention.This study proposes a general framework for analyzing morphology using scanned 3D landmarks, and explores phenotype features identifying population root Japanese archipelago.After registration dense points, we investigate both PCA Mean Hyperplane exploring variations.Then, in order to reduce in-population variance statistical features, normalize them firstly, explore identification...
This paper proposes a novel channel attention method that can adaptively fuse the global pooling features for object detection. The aim is to address learning of suboptimal weights which caused by existing mechanisms using fixed operation. First, feature extraction module proposed simultaneously utilize max and average extracting overall prominent features. Second, an adaptive balance contributions two enhanced operator, obtaining Subsequently, differential operator capture differences...