A5043630267- Human Pose and Action Recognition
- Advanced Neural Network Applications
- 3D Shape Modeling and Analysis
- Advanced Image and Video Retrieval Techniques
- Advanced Vision and Imaging
- Image Retrieval and Classification Techniques
- Hand Gesture Recognition Systems
- Multimodal Machine Learning Applications
- Anomaly Detection Techniques and Applications
- Gait Recognition and Analysis
- 3D Surveying and Cultural Heritage
- Video Surveillance and Tracking Methods
- Computer Graphics and Visualization Techniques
- Domain Adaptation and Few-Shot Learning
- Interactive and Immersive Displays
- Magnetic Properties of Alloys
- Advanced Image Fusion Techniques
- Robotics and Sensor-Based Localization
- Color Science and Applications
- Magnetic and transport properties of perovskites and related materials
- Visual Attention and Saliency Detection
- Image and Signal Denoising Methods
- Video Analysis and Summarization
- Image and Video Quality Assessment
- Rare-earth and actinide compounds
Tongji University
2016-2025
Chinese PLA General Hospital
2024-2025
University of Florence
2020-2024
National Clinical Research
2024
University of Electronic Science and Technology of China
2024
Aerospace Information Research Institute
2023-2024
Chinese Academy of Sciences
2023-2024
Nanjing General Hospital of Nanjing Military Command
2019-2024
Chinese Academy of Medical Sciences & Peking Union Medical College
2024
Beijing Institute of Big Data Research
2023
Recent progresses in salient object detection have exploited the boundary prior, or background information, to assist other saliency cues such as contrast, achieving state-of-the-art results. However, their usage of prior is very simple, fragile, and integration with mostly heuristic. In this work, we present new methods address these issues. First, propose a robust measure, called connectivity. It characterizes spatial layout image regions respect boundaries much more robust. has an...
Regression based methods are not performing as well detection for human pose estimation. A central problem is that the structural information in exploited previous regression methods. In this work, we propose a structure-aware approach. It adopts reparameterized representation using bones instead of joints. exploits joint connection structure to define compositional loss function encodes long range interactions pose. simple, effective, and general both 2D 3D estimation unified setting....
We extends the previous 2D cascaded object pose regression work [9] in two aspects so that it works better for 3D articulated objects. Our first contribution is pose-indexed features generalize parameterized and achieve invariance to transformations. second a principled hierarchical adapted structure. It therefore more accurate faster. Comprehensive experiments verify state-of-the-art accuracy efficiency of proposed approach on challenging hand estimation problem, public dataset our new dataset.
This study aimed to explore the specific efficacy of rituximab (RTX) in treatment membranous nephropathy (MN) and compare analyze differences effectiveness among various regimens, with objective identifying optimal protocol suitable for medical environment China. retrospective focused on patients MN who were treated RTX hospitalized at First Medical Center PLA General Hospital between January 1, 2019, December 30, 2022. These followed up more than one year. We collected clinical data from...
Designing hardware accelerators for deep neural networks (DNNs) has been much desired. Nonetheless, most of these existing are built either convolutional (CNNs) or recurrent (RNNs). Recently, the Transformer model is replacing RNN in natural language processing (NLP) area. However, because intensive matrix computations and complicated data flow being involved, design never reported. In this paper, we propose first accelerator two key components, i.e., multi-head attention (MHA) ResBlock...
Unlike most previous HOI methods that focus on learning better human-object features, we propose a novel and complementary approach called category query learning. Such queries are explicitly associated to interaction categories, converted image specific representation via transformer decoder, learnt an auxiliary image-level classification task. This idea is motivated by earlier multi-label method, but for the first time applied challenging Our method simple, general effective. It validated...
Regression based methods are not performing as well detection for human pose estimation. A central problem is that the structural information in exploited previous regression methods. In this work, we propose a structure-aware approach. It adopts reparameterized representation using bones instead of joints. exploits joint connection structure to define compositional loss function encodes long range interactions pose. simple, effective, and general both 2D 3D estimation unified setting....
In this work, we present a novel and effective framework to facilitate object detection with the instance-level segmentation information that is only supervised by bounding box annotation. Starting from joint instance network, propose recursively estimate pseudo ground-truth masks network training, then enhance top-down feedbacks. The ground truth mask parameters are optimized alternatively mutually benefit each other. To obtain promising in iteration, embed graphical inference incorporates...
Snow profiles can provide reliable and detailed snow parameters for global change research, but it is still challenging to reconstruct large-scale spatiotemporally continuous profiles. Synthetic aperture radar (SAR) tomography (TomoSAR) has been proven be a promising method reconstructing the 3D structure of targets used from ground-based SAR. However, experimental condition SAR ideal algorithms developed have some problems when applied airborne spaceborne sensors, such as large...
Self-assembled iron silicide nanowires were formed by depositing 1ML of Fe onto Si(110) at 700°C in ultrahigh vacuum. The have average dimensions 5nm high ×10nm wide ×μm long, as measured with ex situ atomic force microscopy. High-resolution electron microscopy identifies the crystal structure cubic FeSi2 orientation FeSi2(1¯11)∕∕Si(11¯1), FeSi2⟨110⟩∕∕Si⟨110⟩. Magnetometer measurements show a magnetic moment 0.3Bohr magneton per atom 2K. This property metastable opens up possibility for...
Rapid development of checkpoint inhibitors has provided significant breakthroughs for cancer stem cell (CSC) therapy, while the therapeutic efficacy is restricted by hypoxia-mediated tumor immune evasion, especially hypoxia-induced CD47 overexpression in CSCs. Herein, we developed a genetically engineered CSC membrane-coated hollow manganese dioxide (hMnO2@gCMs) to elicit robust antitumor immunity blocking and alleviating hypoxia ultimately achieve eradication The hMnO2 core effectively...
Hierarchical segmentation based object proposal methods have become an important step in modern detection paradigm. However, standard single-way hierarchical are fundamentally flawed that the errors early steps cannot be corrected and accumulate. In this work, we propose a novel multi-branch approach alleviates such problems by learning multiple merging strategies each complementary manner, one strategy could others. Our achieves state-of-the-art performance for both tasks, comparing to...
The abundance of remote sensing imagery available has been extensively used for mapping snow cover extent in mountainous regions. However, previous studies have paid little attention to quantifying the uncertainties inherent algorithms when using Landsat observations, particularly context delineating snowline—a pivotal parameter understanding spatiotemporal dynamics cover. Additionally, there is an urgent need automated processing approach capable monitoring alpine snowline across expansive...
Recently, sketch-based 3D shape retrieval has received growing attention in the community of computer graphics and vision. Most previous works focus on problem how to reduce large cross-modality difference between 2D sketch data make significant progress. Nevertheless, little been paid another important deal with noise data. For first time, this work investigates noisy It firstly provides qualitative insightful analysis impact noise, revealing that are a key factor for unsatisfactory...
Visual relationships are crucial for visual perception and reasoning, cover tasks like Scene Graph Generation, Human-Object Interaction, object affordance. Despite significant efforts, this field still suffers from the following limitations: specialists a specific task without considering similar ones, strict complex formulations with limited flexibility, underexploited reasoning language knowledge. To solve these limitations, we seek to build new framework, one model all tasks, over Large...