A5017829201- Advanced Neural Network Applications
- Advanced Image and Video Retrieval Techniques
- Domain Adaptation and Few-Shot Learning
- Head and Neck Cancer Studies
- Thyroid Cancer Diagnosis and Treatment
- Video Surveillance and Tracking Methods
- Autonomous Vehicle Technology and Safety
- Advanced Vision and Imaging
- Radiomics and Machine Learning in Medical Imaging
- Fuel Cells and Related Materials
- Meteorological Phenomena and Simulations
- Robotics and Sensor-Based Localization
- Precipitation Measurement and Analysis
- Multimodal Machine Learning Applications
- Machine Learning and ELM
- Electrocatalysts for Energy Conversion
- Traditional and Medicinal Uses of Annonaceae
- Chromatography in Natural Products
- Phytochemistry and Biological Activities
- Traditional Chinese Medicine Analysis
- Advanced Graph Neural Networks
- Gaze Tracking and Assistive Technology
- Proteins in Food Systems
- Graph Theory and Algorithms
- Image Processing Techniques and Applications
Jiangnan University
2021-2025
Peking University First Hospital
2025
Tianjin Normal University
2025
Peking University
2025
Yunnan University
2025
Fujian University of Traditional Chinese Medicine
2010-2024
Chongqing University of Technology
2011-2024
Chinese University of Hong Kong
2024
Chinese Academy of Medical Sciences & Peking Union Medical College
2019-2024
Sichuan Agricultural University
2022-2024
Despite deep neural networks have demonstrated extraordinary power in various applications, their superior performances are at expense of high storage and computational costs. Consequently, the acceleration compression attracted much attention recently. Knowledge Transfer (KT), which aims training a smaller student network by transferring knowledge from larger teacher model, is one popular solutions. In this paper, we propose novel transfer method treating it as distribution matching...
While general object detection with deep learning has achieved great success in the past few years, performance and efficiency of detecting small objects are far from satisfactory. The most common effective way to promote is use high-resolution images or feature maps. However, both approaches induce costly computation since computational cost grows squarely as size features increases. To get best two worlds, we propose QueryDet that uses a novel query mechanism accelerate inference speed...
Data association across frames is at the core of Multiple Object Tracking (MOT) task. This problem usually solved by a traditional graph-based optimization or directly learned via deep learning. Despite their popularity, we find some points worth studying in current paradigm: 1) Existing methods mostly ignore context information among tracklets and intra-frame detections, which makes tracker hard to survive challenging cases like severe occlusion. 2) The end-to-end solely rely on data...
Monocular 3D lane detection is a challenging task due to its lack of depth information. A popular solution first transform the front-viewed (FV) images or features into bird-eye-view (BEV) space with inverse perspective mapping (IPM) and detect lanes from BEV features. However, reliance IPM on flat ground assumption loss context information make it inaccurate restore representations. An attempt has been made get rid predict FV representations directly, while still underperforms other...
Currently prevalent multi-modal 3D detection methods rely on dense detectors that usually use Bird's-Eye-View (BEV) feature maps. However, the cost of such BEV maps is quadratic to range, making it not scalable for long-range detection. Recently, LiDAR-only fully sparse architecture has been gaining attention its high efficiency in perception. In this paper, we study how develop a detector. Specifically, our proposed detector integrates well-studied 2D instance segmentation into LiDAR side,...
Drowsiness poses a serious challenge to road safety and various in-cabin sensing technologies have been experimented with monitor driver alertness. Cameras offer convenient means for contactless sensing, but they may violate user privacy require complex algorithms accommodate (e.g., sunglasses) environmental lighting conditions) constraints. This paper presents lightweight convolution neural network that measures eye closure based on images captured by wearable glass prototype, which...
Quantum dots (QDs), as unique nanoparticle probes, have been used in vivo fluorescence imaging such cancers. Due to the novel characteristics fluorescence, QDs represent a family of promising substances be experimental and clinical imaging. Thus far, toxicity harmful health effects from exposure (including environmental exposure) are not recognized, but largely concerned by public. To assess biological QDs, we established mouse model acute chronic QDs. Results present study suggested that QD...
Recently neural architecture search (NAS) has raised great interest in both academia and industry. However, it remains challenging because of its huge non-continuous space. Instead applying evolutionary algorithm or reinforcement learning as previous works, this paper proposes a direct sparse optimization NAS (DSO-NAS) method. The motivation behind DSO-NAS is to address the task view model pruning. To achieve goal, we start from completely connected block, then introduce scaling factors...
Abstract. Deriving large-scale and high-quality precipitation products from satellite remote-sensing spectral data is always challenging in quantitative estimation (QPE), limited studies have been conducted even using China's latest Fengyun-4A (FY-4A) geostationary satellite. Taking three rainstorm events over South China as examples, a machine-learning-based regression model was established the random forest (RF) method to derive QPE FY-4A observations, conjunction with cloud parameters...
Data augmentation has been an indispensable tool to improve the performance of deep neural networks, however can hardly transfer among different tasks and datasets. Consequently, a recent trend is adopt AutoML technique learn proper policy without extensive hand-crafted tuning. In this paper, we propose efficient differentiable search algorithm called Direct Differentiable Augmentation Search (DDAS). It exploits meta-learning with one-step gradient update continuous relaxation expected...
Cold storage preserves lemon fruit quality; however, it can result in significant chilling injury (CI). The effects of pre- and post-harvest methyl jasmonate (MeJA) treatments at four concentrations (0, 0.1, 0.3, 0.5 mM) on CI sensory quality lemons during 80 d 7-10 °C were investigated. Both MeJA reduced CI, weight loss (WL) maintained higher firmness, total soluble solids (TSS), acidity (TA) than the controls. Antioxidant enzyme activities decreased control but increased both MeJA-treated...
3D object detection from multi-view images has drawn much attention over the past few years. Existing methods mainly establish representations and adopt a dense head for detection, or employ queries distributed in space to localize objects. In this paper, we design Multi-View 2D Objects guided Object Detector (MV2D), which can lift any detector detection. Since detections provide valuable priors existence, MV2D exploits detectors generate conditioned on rich image semantics. These...
Recently Neural Architecture Search (NAS) has aroused great interest in both academia and industry, however it remains challenging because of its huge non-continuous search space. Instead applying evolutionary algorithm or reinforcement learning as previous works, this paper proposes a Direct Sparse Optimization NAS (DSO-NAS) method. In DSO-NAS, we provide novel model pruning view to problem. specific, start from completely connected block, then introduce scaling factors scale the...
Abstract Introduction Studies have revealed that age is associated with the risk of lateral lymph node metastasis (LLNM) in papillary thyroid cancer (PTC). This study aimed to identify optimal cut point for a more precise prediction model LLNM and reveal differences factors between patients distinct stages. Methods A total 499 who had undergone thyroidectomy neck dissection (LND) PTC were enrolled. The locally weighted scatterplot smoothing (LOWESS) curve ‘changepoint’ package used using R....