A5081011415- Ultrasound Imaging and Elastography
- Topic Modeling
- Natural Language Processing Techniques
- Advanced Neural Network Applications
- Mosquito-borne diseases and control
- Ultrasonics and Acoustic Wave Propagation
- Viral Infections and Vectors
- Multimodal Machine Learning Applications
- Plasma Applications and Diagnostics
- Metaheuristic Optimization Algorithms Research
- Sparse and Compressive Sensing Techniques
- Domain Adaptation and Few-Shot Learning
- Radiomics and Machine Learning in Medical Imaging
- Indoor and Outdoor Localization Technologies
- Energy Efficient Wireless Sensor Networks
- Photoacoustic and Ultrasonic Imaging
- AI in cancer detection
- Esophageal Cancer Research and Treatment
- Advanced Image and Video Retrieval Techniques
- Plasma Diagnostics and Applications
- Neural Networks and Applications
- Dengue and Mosquito Control Research
- Viral gastroenteritis research and epidemiology
- Bone Tissue Engineering Materials
- Viral Infections and Immunology Research
Sun Yat-sen University Cancer Center
2014-2025
Sun Yat-sen University
2014-2025
South China University of Technology
2015-2024
Ministry of Education of the People's Republic of China
2012-2024
Renmin University of China
2024
Jimei University
2012-2024
Zhengzhou University
2024
Henan Polytechnic University
2023-2024
Xi'an University of Architecture and Technology
2011-2023
Southwest Jiaotong University
2001-2023
Object detection with transformers (DETR) reaches competitive performance Faster R-CNN via a transformer encoder-decoder architecture. Inspired by the great success of pre-training in natural language processing, we propose pretext task named random query patch to Unsupervisedly Pre-train DETR (UP-DETR) for object detection. Specifically, randomly crop patches from given image and then feed them as queries decoder. The model is pre-trained detect these original image. During pre-training,...
Hongbo Zhang, Junying Chen, Feng Jiang, Fei Yu, Zhihong Guiming Jianquan Li, Xiangbo Wu, Zhang Zhiyi, Qingying Xiao, Xiang Wan, Benyou Wang, Haizhou Li. Findings of the Association for Computational Linguistics: EMNLP 2023.
Visual contexts often help to recognize named entities more precisely in short texts such as tweets or snapchat. For example, one can identify "Charlie'' a name of dog according the user posts. Previous works on multimodal entity recognition ignore corresponding relations visual objects and entities. are considered fine-grained image representations. sentence with multiple types, relevant be utilized capture different information. In this paper, we propose neural network which combines...
Visual question answering aims to answer the natural language about a given image. Existing graph-based methods only focus on relations between objects in an image and neglect importance of syntactic dependency words question. To simultaneously capture question, we propose novel dual channel graph convolutional network (DC-GCN) for better combining visual textual advantages. The DC-GCN model consists three parts: I-GCN module image, Q-GCN attention alignment align representations...
Convolutional neural network (CNN) architectures are generally heavy on memory and computational requirements which make them infeasible for embedded systems with limited hardware resources. We propose dual convolutional kernels (DualConv) constructing lightweight deep networks. DualConv combines 3×3 1×1 to process the same input feature map channels simultaneously exploits group convolution technique efficiently arrange filters. can be employed in any CNN model such as VGG-16 ResNet-50...
The process of full-thickness skin regeneration is complex and has many parameters involved, which makes it difficult to use a single dressing meet the various requirements complete at same time. Therefore, developing hydrogel dressings with multifunction, including tunable rheological properties aperture, hemostatic, antibacterial super cytocompatibility, desirable candidate in wound healing. In this study, series hydrogels were developed via hydrogen bond covalent between chitosan (CS)...
There is an urgent need to develop a series of multifunctional materials with good biocompatibility, high mechanical strength, hemostatic properties, antiadhesion, and anti-infection for applications in wound care. However, successfully developing challenging. In this study, two superhydrophobic composite coatings strong antifouling blood repellency, fast hemostasis, antibacterial activity are prepared on cotton fabric surface by simple, green, low-cost one-step dip-coating technology. The...
Medical ultrasound imaging stands out from other modalities in providing real-time diagnostic capability at an affordable price while being physically portable. This article explores the suitability of using GPUs as primary signal and image processors for future medical systems. A case study on synthetic aperture (SA) illustrates promise high-performance such
Antibody dependent enhancement (ADE) of dengue virus (DENV) infection is identified as the main risk factor severe Dengue diseases. Through opsonization by subneutralizing or non-neutralizing antibodies, DENV suppresses innate cell immunity to facilitate viral replication. However, it largely unknown whether suppression type-I IFN necessary for a successful ADE infection. Here, we report that both and DENV-ADE induce an early ISG (NOS2) expression through RLR-MAVS signalling axis independent...
Background: Video-assisted thoracoscopic surgery has been identified as priori choice compared with open approaches in esophageal cancer surgery. With the developments Da Vinci robotic system, robot-assisted minimally invasive esophagectomy (RAMIE) increasingly popular. However, whether RAMIE could be a better over thoraco-laparoscopic (TLMIE) is unclear. Methods: The clinicopathological characteristics of patients who received or TLMIE modern two-field lymph node dissection Sun Yat-sen...
In this paper, we summarize the 1st NTIRE challenge on stereo image super-resolution (restoration of rich details in a pair low-resolution images) with focus new solutions and results. This has 1 track aiming at problem under standard bicubic degradation. total, 238 participants were successfully registered, 21 teams competed final testing phase. Among those participants, 20 submitted results PSNR (RGB) scores better than baseline. establishes benchmark for SR.
This paper reviews the NTIRE 2022 Challenge on Super-Resolution and Quality Enhancement of Compressed Video. In this challenge, we proposed LDV 2.0 dataset, which includes dataset (240 videos) 95 additional videos. challenge three tracks. Track 1 aims at enhancing videos compressed by HEVC a fixed QP. 2 3 target both super-resolution quality enhancement video. They require x2 x4 super-resolution, respectively. The tracks totally attract more than 600 registrations. test phase, 8 teams, teams...
DEtection TRansformer (DETR) for object detection reaches competitive performance compared with Faster R-CNN via a transformer encoder-decoder architecture. However, trained scratch transformers, DETR needs large-scale training data and an extreme long schedule even on COCO dataset. Inspired by the great success of pre-training transformers in natural language processing, we propose novel pretext task named random query patch Unsupervised Pre-training (UP-DETR). Specifically, randomly crop...
DEtection TRansformer (DETR) for object detection reaches competitive performance compared with Faster R-CNN via a transformer encoder-decoder architecture. However, trained scratch transformers, DETR needs large-scale training data and an extreme long schedule even on COCO dataset. Inspired by the great success of pre-training transformers in natural language processing, we propose novel pretext task named random query patch Unsupervised Pre-training (UP-DETR). Specifically, randomly crop...
Adapting a language model into specific domain, a.k.a `domain adaption', is common practice when specialized knowledge, e.g. medicine, not encapsulated in general like Llama2. The challenge lies the heterogeneity of data across two training stages, as it varies languages, genres, or formats. To tackle this and simplify learning protocol, we propose to transform heterogeneous data, from both pre-training supervised unified, simple input-output pair format. We validate new protocol domains...