A5101758971- Human Pose and Action Recognition
- Computational and Text Analysis Methods
- Domain Adaptation and Few-Shot Learning
- Image and Signal Denoising Methods
- AI in cancer detection
- Multimodal Machine Learning Applications
- Advanced Neural Network Applications
- Industrial Vision Systems and Defect Detection
- Advanced Image Fusion Techniques
- Vehicle License Plate Recognition
Shandong Jianzhu University
2022-2023
Lanzhou Jiaotong University
2022
First Hospital of Jilin University
2019
Jilin University
2019
Text classification is the most common application of Natural Language Processing(NLP), and Transformer models have dominated field in recent years. Currently, pre-training modeling text through deep learning methods a way classification. This paper firstly proposes an improved XLNet model based on problems long-term dependence insufficient contextual semantic expression previous pre-trained language models, uses to represent as low-dimensional word vectors obtain sequences. Secondly,...
Video question answering is an increasingly vital research field, spurred by the rapid proliferation of video content online and urgent need for intelligent systems that can comprehend interact with this content. Existing methodologies often lean towards understanding cross-modal information interaction modeling but tend to overlook crucial aspect comprehensive understanding. To address gap, we introduce multi-modal multi-layer enhancement network, a groundbreaking framework emphasizing...
Abstract In this article, a new optimized method for diagnosing and analyzing breast cancer from the mammography images is presented. regard, preprocessing used to remove Gaussian noises that are happen in also additional areas. Then, image segmentation performed on determine areas where contrast material perceptible. Afterward, combined feature extraction based discrete wavelet transform gray‐level co‐occurrence matrix proposed extracting important information fromthe images. Finally,...
VCIP 2022 "Tire pattern image classification based on lightweight network challenge" aims to design networks that correctly classify tire surface tread patterns and indentation images using less overhead. To this end, we present a novel network. Concretely, adopt the ShuffleNet-V2-x0.5 as our backbone. reduce computation complexity, introduce Space-To-Depth Anti-Alias Downsampling modules pre-process input image. Moreover, enhance ability of model, knowledge distillation strategy by...