A5070709298- Human Pose and Action Recognition
- Human Motion and Animation
- 3D Shape Modeling and Analysis
- Video Analysis and Summarization
- Hand Gesture Recognition Systems
- Virtual Reality Applications and Impacts
- Robot Manipulation and Learning
- Gene expression and cancer classification
- Renal Diseases and Glomerulopathies
- Advanced Vision and Imaging
- Functional Brain Connectivity Studies
- Olfactory and Sensory Function Studies
- Neural dynamics and brain function
- Tea Polyphenols and Effects
- Advanced Bandit Algorithms Research
- Bioinformatics and Genomic Networks
- Generative Adversarial Networks and Image Synthesis
- Advanced Neuroimaging Techniques and Applications
- Nanocluster Synthesis and Applications
- Advanced Memory and Neural Computing
- Tactile and Sensory Interactions
- Ferroelectric and Negative Capacitance Devices
- Photonic and Optical Devices
- Domain Adaptation and Few-Shot Learning
- Computational Drug Discovery Methods
Chongqing University
2025
Fujian Institute of Research on the Structure of Matter
2024
Chinese Academy of Sciences
2024
META Health
2022-2024
China Pharmaceutical University
2020-2024
ETH Zurich
2024
Jiangsu University
2024
Shenzhen University
2019-2023
Technical University of Denmark
2023
Institute of Psychology, Chinese Academy of Sciences
2022-2023
Over the past few years, Spiking Neural Networks (SNNs) have become popular as a possible pathway to enable low-power event-driven neuromorphic hardware. However, their application in machine learning largely been limited very shallow neural network architectures for simple problems. In this paper, we propose novel algorithmic technique generating an SNN with deep architecture, and demonstrate its effectiveness on complex visual recognition problems such CIFAR-10 ImageNet. Our applies both...
We introduce a real-time and calibration-free facial performance capture framework based on sensor with video depth input. In this framework, we develop an adaptive PCA model using shape correctives that adjust on-the-fly to the actor's expressions through incremental PCA-based learning. Since fitting of progressively improves during performance, do not require extra or training session build model. As result, system is highly deployable easy use: it can faithfully track any individual,...
We present a system for real-time hand-tracking to drive virtual and augmented reality (VR/AR) experiences. Using four fisheye monochrome cameras, our generates accurate low-jitter 3D hand motion across large working volume diverse set of users. achieve this by proposing neural network architectures detecting hands estimating keypoint locations. Our detection robustly handles variety real world environments. The estimation leverages tracking history produce spatially temporally consistent...
Real-time tracking of human body motion is crucial for interactive and immersive experiences in AR/VR. However, very limited sensor data about the available from standalone wearable devices such as HMDs (Head Mounted Devices) or AR glasses. In this work, we present a reinforcement learning framework that takes sparse signals an HMD two controllers, simulates plausible physically valid full motions. Using high quality dense supervision during training, simple policy network can learn to...
Optical marker-based motion capture is the dominant way for obtaining high-fidelity human body animation special effects, movies, and video games. However, has seen limited application to hand due difficulty of automatically identifying (or labeling) identical markers on self-similar fingers. We propose a technique that frames labeling problem as keypoint regression conducive solution using convolutional neural networks. demonstrate robustness our occlusion, ghost markers, shape, even...
Capturing human activities that involve both gross full-body motion and detailed hand manipulation of objects is challenging for standard capture systems. We introduce a new method creating natural scenes with such activities. The input to our includes motions the acquired simultaneously by system. Our then automatically synthesizes physically plausible can seamlessly integrate motions. Instead producing one "optimal" solution, presents set exploit wide variety strategies. propose randomized...
Abstract Toll-like receptor 9 (TLR9) senses bacterial DNA characteristic of unmethylated CpG motifs to induce innate immune response. TLR9 is de novo expressed in podocytes some patients with glomerular diseases, but its role podocyte injury remains undetermined. Since activates p38 MAPK and NFkB that are known mediate apoptosis, we hypothesized induces apoptosis diseases. We treated immortalized puromycin aminonucleosides (PAN) observed accompanied by upregulation. Prevention upregulation...
Background: Schizophrenia (SZ), bipolar disorder (BD), and major depressive (MDD) are distinct diagnostic categories in current psychiatric nosology, yet there is increasing evidence for shared clinical biological features these disorders. No previous studies have examined brain structural concurrently 3 The aim of this study was to identify the extent alterations SZ, BD, MDD. We gray matter (GM) volume white (WM) integrity a total 485 individuals (135 with 86 108 MDD, 156 healthy controls...
Real-time human motion reconstruction from a sparse set of (e.g. six) wearable IMUs provides non-intrusive and economic approach to capture. Without the ability acquire position information directly IMUs, recent works took data-driven approaches that utilize large datasets tackle this under-determined problem. Still, challenges remain such as temporal consistency, drifting global joint motions, diverse coverage types on various terrains. We propose novel method simultaneously estimate...
Abstract Acquired drug resistance is one of the most common limitations for clinical response colon cancer to 5-Fluorouracil (5-FU)-based chemotherapy. The relevant molecular mechanisms might be diversity, but still not elucidated clearly. In this study, we aimed investigate potential c-Fos, a subfamily activator protein-1, in 5-FU chemoresistance. We determined that phosphorylated c-Fos promoted cells by facilitating stemness. Mechanically, treatment induced autolysosome-dependent...
The leaching dynamics of the Ag–S interface and mechanism CO 2 electroreduction for Ag 25 (SR) 18 nanocluster differ significantly from those its Au analogue.
We present a physics-based approach to synthesizing motion of virtual character in dynamically varying environment. Our views the responsive as sequence solutions constrained optimization problem formulated at every time step. This framework allows programmer specify active control strategies using intuitive kinematic goals, significantly reducing engineering effort entailed body control. can incorporate changes character's surroundings through synthetic visual sensory system and create...
Real-time adaptation of a motion capture sequence to virtual environments with physical perturbations requires robust control strategies. This paper describes an optimal feedback controller for tracking that allows on-the-fly re-planning long-term goals and adjustments in the final completion time. We first solve offline trajectory problem abstract dynamic model captures essential relation between contact forces momenta. A policy is then derived used simulate online. Simulation results...
Anopheles sinensis is an important mosquito vector of Plasmodium vivax, which the most frequent and widely distributed cause recurring malaria throughout Asia, particularly in China, Korea, Japan.We performed 454 next-generation sequencing obtained a draft sequence A. assembled into scaffolds spanning 220.8 million base pairs. Analysis this genome sequence, we observed expansion contraction several immune-related gene families anopheline relative to culicine species. These differences...
Natural hand manipulations exhibit complex finger maneuvers adaptive to object shapes and the tasks at hand. Learning dexterous manipulation from data in a brute force way would require prohibitive amount of examples effectively cover combinatorial space 3D activities. In this paper, we propose hand-object spatial representation that can achieve generalization limited data. Our combines global shape as voxel occupancies with local geometric details samples closest distances. This is used by...
Most commercial virtual reality applications with self avatars provide users a "one-size fits all" avatar. While the height of this body may be scaled to user's height, other proportions, such as limb length and hand size, are rarely customized fit an individual user. Prior research has shown that mismatches between users' their actual bodies can affect size perception feelings ownership. In paper, we consider how concepts related illusion, user experience, task efficiency influenced by...
Learning latent representations of registered meshes is useful for many 3D tasks. Techniques have recently shifted to neural mesh autoencoders. Although they demonstrate higher precision than traditional methods, remain unable capture fine-grained deformations. Furthermore, these methods can only be applied a template-specific surface mesh, and not applicable more general meshes, like tetrahedrons non-manifold meshes. While graph convolution employed, lack performance in reconstruction...
As one of the most common malignancies worldwide, Hepatocellular carcinoma (HCC) has been treated by Sorafenib, which is first approved target drug FDA for advanced HCC. However, resistance obstacles to its application. a typical characteristic solid tumors, hypoxia become key cause chemotherapy and radiotherapy. It important elucidate underlying mechanisms Sorafenib under hypoxia. In this study, morphological changes hepatocellular cells were observed Live Cell Imaging System Transmission...
Replicating a user's pose from only wearable sensors is important for many AR/VR applications. Most existing methods motion tracking avoid environment interaction apart foot-floor contact due to their complex dynamics and hard constraints. However, in daily life people regularly interact with environment, e.g. by sitting on couch or leaning desk. Using Reinforcement Learning, we show that headset controller pose, if combined physics simulation observations can generate realistic full-body...
We present a novel motion prior, called PhaseMP, modeling probability distribution on pose transitions conditioned by frequency domain feature extracted from periodic autoencoder. The phase further enforces the to be unidirectional (i.e. no backward movement in time), which more stable and natural motions can generated. Specifically, our prior useful for accurately estimating 3D human presence of challenging input data, including long periods spatial temporal occlusion, as well noisy sensor...