A5106084673- Particle physics theoretical and experimental studies
- Quantum Chromodynamics and Particle Interactions
- High-Energy Particle Collisions Research
- Space Satellite Systems and Control
- Video Surveillance and Tracking Methods
- Advanced Image Processing Techniques
- Image Enhancement Techniques
- Generative Adversarial Networks and Image Synthesis
- Astro and Planetary Science
- Dark Matter and Cosmic Phenomena
- Indoor and Outdoor Localization Technologies
- Advanced MRI Techniques and Applications
- Nuclear physics research studies
- Black Holes and Theoretical Physics
- Monoclonal and Polyclonal Antibodies Research
- Thyroid Cancer Diagnosis and Treatment
- Face recognition and analysis
- Face and Expression Recognition
- Underwater Vehicles and Communication Systems
- Medical Image Segmentation Techniques
- Multimodal Machine Learning Applications
- Advanced Vision and Imaging
- Topic Modeling
- Spacecraft Dynamics and Control
- Soft Robotics and Applications
Peking University
2020-2025
University of Science and Technology of China
2013-2025
Tongji University
2013-2025
Fudan University
2020-2025
Shanghai Jiao Tong University
2020-2025
Ruijin Hospital
2016-2025
Obstetrics and Gynecology Hospital of Fudan University
2024
People's Government of Yunnan Province
2024
Yunnan University
2024
Memorial University of Newfoundland
2020-2024
On January 20, 2020, a new coronavirus epidemic with human-to-human transmission was officially declared by the Chinese government, which caused significant public panic in China. In light of disease 2019 outbreak, pregnant women may be particularly vulnerable and special need for preventive mental health strategies. Thus far, no reports exist to investigate response outbreak.This study aimed examine impact outbreak on prevalence depressive anxiety symptoms corresponding risk factors among...
We present a novel face alignment framework based on coarse-to-fine shape searching. Unlike the conventional cascaded regression approaches that start with an initial and refine in manner, our approach begins coarse search over space contains diverse shapes, employs solution to constrain subsequent finer of shapes. The unique stage-by-stage progressive adaptive i) prevents final from being trapped local optima due poor initialisation, common problem encountered by approaches; ii) improves...
Recent developed deep unsupervised methods allow us to jointly learn representation and cluster unlabelled data. These clustering %like DAC start with mainly focus on the correlation among samples, e.g., selecting high precision pairs gradually tune feature representation, which neglects other useful correlations. In this paper, we propose a novel framework, named comprehensive mining~(DCCM), for exploring taking full advantage of various kinds correlations behind unlabeled data from three...
How to optimize ranking metrics such as Normalized Discounted Cumulative Gain (NDCG) is an important but challenging problem, because are either flat or discontinuous everywhere, which makes them hard be optimized directly. Among existing approaches, LambdaRank a novel algorithm that incorporates into its learning procedure. Though empirically effective, it still lacks theoretical justification. For example, the underlying loss optimizes for remains unknown until now. Due this, there no...
The human hearing range is from 20 Hz to kHz. However, many animals can hear much higher sound frequencies. Dolphins, especially, have a up 300 To our knowledge, there no data of reported wide-band frequency earphone satisfy both humans and animals. Here, we show that graphene earphones, packaged into commercial casings play sounds ranging 100 50 By using one-step laser scribing technology, wafer-scale flexible earphones be obtained in 25 min. Compared with normal earphone, the has wider...
Unsupervised image-to-image translation aims at learning a mapping between two visual domains. However, across large geometry variations al- ways ends up with failure. In this work, we present novel disentangle-and-translate framework to tackle the complex objects task. Instead of on image space directly, disentangle into Cartesian product appearance and latent spaces. Specifically, first in- troduce prior loss conditional VAE encourage network learn independent but com- plementary...
Hi-C is commonly used to study three-dimensional genome organization. However, due the high sequencing cost and technical constraints, resolution of most datasets coarse, resulting in a loss information biological interpretability. Here we develop DeepHiC, generative adversarial network, predict high-resolution contact maps from low-coverage data. We demonstrated that DeepHiC capable reproducing data as few 1% downsampled reads. Empowered by training, our method can restore fine-grained...
Computer vision field has achieved great success in interpreting semantic meanings from images, yet its algorithms can be brittle for tasks with adverse conditions and the ones suffering data/label pair limitation. Among these is in-bed human pose monitoring significant value many healthcare applications. In-bed natural settings involves estimation complete darkness or full occlusion. The lack of publicly available datasets hinders applicability successful this task. In paper, we introduce...
Ischemic heart disease is one of the leading causes mortality worldwide, and thus calls for development more effective therapeutic strategies. This study aimed to identify potential targets coronary (CHD) myocardial infarction (MI) by investigating causal relationship between plasma proteins these conditions.
ABSTRACT Vaccines have been the primary remedy in global fight against coronavirus disease 2019 (COVID-19). The receptor-binding domain (RBD) of spike protein, a critical viral immunogen, is affected by heterogeneity its glycan structures and relatively low immunogenicity. Here, we describe scalable synthetic platform that enables precise synthesis homogeneously glycosylated RBD, facilitating elucidation carbohydrate structure–function relationships. Five RBDs bearing biantennary glycans...
Learning-to-Rank deals with maximizing the utility of a list examples presented to user, items higher relevance being prioritized. It has several practical applications such as large-scale search, recommender systems, document summarization and question answering. While there is widespread support for classification regression based learning, learning-to-rank in deep learning been limited. We introduce TensorFlow Ranking, first open source library solving ranking problems framework. highly...
Recent advances on deep learning models come at the price of formidable training cost. The increasing model size is one root causes, but another less-emphasized fact that data scale actually a similar speed as scale, and cost proportional to both them. Compared rapidly evolving architecture, how efficiently use (especially for expensive foundation pretraining) less explored difficult realize due lack convenient framework focus efficiency capabilities. To this end, we present DeepSpeed Data...
In order to study the developing process of surface discharges initiated by immobilized metallic particles attached Gas Insulated Substation (GIS) insulators under gradually-raised applying voltages, long-term laboratory tests were conducted on a well-established 220 kV test assembly observe entire evaluation from its inception gradual expansion and eventual occurrence flashover. Real-time detection was record conventional impulse current (CIC) signals, UHF ultrasonic signals light emission...
Lane detection plays a key role in autonomous driving. While car cameras always take streaming videos on the way, current lane works mainly focus individual images (frames) by ignoring dynamics along video. In this work, we collect new video instance (VIL-100) dataset, which contains 100 with total 10,000 frames, acquired from different real traffic scenarios. All frames each are manually annotated to high-quality instance-level annotation, and set of frame-level video-level metrics included...
Abstract The non-holonomic characteristic of a free-floating space robotic system is used to plan the path manipulator joints, by whose motion base attitude and inertial pose (the position orientation with respect frame) end-effector attain desired values. First, kinematic equations robot are simplified state variables transformed another form composed joint angles. Then, trajectories parameterized using sinusoidal functions, arguments seven-order polynomials. Third, planning problem an...
The improved Kelson-Garvey (ImKG) mass relations are proposed from the differences of mirror nuclei. masses 31 measured proton-rich nuclei with $7\ensuremath{\le}A\ensuremath{\le}41$ and $\ensuremath{-}5\ensuremath{\le}(N\ensuremath{-}Z)\ensuremath{\le}\ensuremath{-}3$ can be remarkably well reproduced by using relations, a root-mean-square deviation 0.398 MeV, which is much smaller than results (0.502 MeV) isobar-mirror (0.647 MeV). This because many more participating involved in ImKG for...
Supervised relation extraction methods based on deep neural network play an important role in the recent information field. However, at present, their performance still fails to reach a good level due existence of complicated relations. On other hand, recently proposed pre-trained language models (PLMs) have achieved great success multiple tasks natural processing through fine-tuning when combined with model downstream tasks. original standard PLM do not include task yet. We believe that...
The dynamics of the charge equilibration (CE) and effects on production neutron-rich isotopes around $N=126$ in multinucleon transfer reactions are still not well understood. In this work, we investigate mechanism CE from different viewpoints by using extended version dinuclear system model (DNS-sysu) improved quantum molecular (ImQMD) model. From macroscopic microscopic dynamical viewpoints, find incomplete for mass asymmetry reaction systems even very deep collisions, behavior ``inverse...
Even though the collaboration between traditional and neuromorphic event cameras brings prosperity to frame-event based vision applications, performance is still confined by resolution gap crossing two modalities in both spatial temporal domains. This paper devoted bridging increasing for images, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i.e.</i> , motion deblurring, events, super-resolving, respectively. To this end, we introduce...