A5100415330- Face recognition and analysis
- 3D Shape Modeling and Analysis
- Generative Adversarial Networks and Image Synthesis
- Human Pose and Action Recognition
- Advanced Vision and Imaging
- Advanced Image Processing Techniques
- Computer Graphics and Visualization Techniques
- Video Surveillance and Tracking Methods
- Insect and Pesticide Research
- Chaos-based Image/Signal Encryption
- Insect Pest Control Strategies
- User Authentication and Security Systems
- Anomaly Detection Techniques and Applications
- Optical measurement and interference techniques
- Winter Sports Injuries and Performance
- Bayesian Methods and Mixture Models
- Image and Video Stabilization
- Neural Networks and Applications
- Gait Recognition and Analysis
- Neurobiology and Insect Physiology Research
- Face and Expression Recognition
- Cryptographic Implementations and Security
- Advanced Numerical Analysis Techniques
- Chaos control and synchronization
- Mosquito-borne diseases and control
Michigan State University
2019-2024
Michigan United
2021
Xi'an Jiaotong University
2020
University of Wisconsin–Madison
2009
Embedding 3D morphable basis functions into deep neural networks opens great potential for models with better representation power. However, to faithfully learn those from an image collection, it requires strong regularization overcome ambiguities involved in the learning process. This critically prevents us high fidelity face which are needed represent images level of details. To address this problem, paper presents a novel approach additional proxies as means side-step regularizations,...
Generating synthetic datasets for training face recognition models is challenging because dataset generation entails more than creating high fidelity images. It involves generating multiple images of same subjects under different factors (e.g., variations in pose, illumination, expression, aging and occlusion) which follows the real image conditional distribution. Previous works have studied using GAN or 3D models. In this work, we approach problem from aspect combining subject appearance...
Gait, the walking pattern of individuals, is one important biometrics modalities. Most existing gait recognition methods take silhouettes or articulated body models as features. These suffer from degraded performance when handling confounding variables, such clothing, carrying and viewing angle. To remedy this issue, we propose a novel AutoEncoder framework, GaitNet, to explicitly disentangle appearance, canonical pose features RGB imagery. The LSTM integrates over time dynamic feature while...
We describe a technique that transforms video from hand-held camera so it appears as if were taken with directed motion. Our method adjusts the to appear nearby viewpoints, allowing 3D movements be simulated. By aiming only for perceptual plausibility, rather than accurate reconstruction, we are able develop algorithms can effectively recreate dynamic scenes single source video. first recovers original motion and sparse set of 3D, static scene points using an off-the-shelf...
Whole-body biometric recognition is an important area of research due to its vast applications in law enforcement, border security, and surveillance. This paper presents the end-to-end design, development evaluation FarSight, innovative software system designed for whole-body (fusion face, gait body shape) recognition. FarSight accepts videos from elevated platforms drones as input outputs a candidate list identities gallery. The address several challenges, including (i) low-quality imagery,...
Traditional 3D face models learn a latent representation of faces using linear subspaces from limited scans single database. The main roadblock building large-scale model diverse databases lies in the lack dense correspondence among raw scans. To address these problems, this paper proposes an innovative framework to jointly nonlinear set scan and establish point-to-point their Specifically, by treating input as unorganized point clouds, we explore use PointNet architectures for converting...
Background Volatile pyrethroid insecticides, such as transfluthrin, have received increasing attention for their potent repellent activities in recent years controlling human disease vectors. It has been long understood that pyrethroids kill insects by promoting activation and inhibiting inactivation of voltage-gated sodium channels. However, the mechanism repellency remains poorly controversial. Methodology/Principal findings Here, we show transfluthrin repels Aedes aegypti a hand-in-cage...
Use of pyrethroid insecticides is a pivotal strategy for mosquito control globally. Commonly known their insecticidal activity by acting on voltage-gated sodium channels, pyrethroids, such as bioallethrin and transfluthrin, are used in coils, emanators other vaporizers to repel mosquitoes biting arthropods. However, whether specific olfactory receptor neurons activated pyrethroids trigger spatial repellency remains unknown.We behavioral electrophysiological approaches elucidate the mechanism...
Inferring 3D structure of a generic object from 2D image is long-standing objective computer vision. Conventional approaches either learn completely CADgenerated synthetic data, which have difficulty in inference real images, or generate 2.5D depth via intrinsic decomposition, limited compared to the full reconstruction. One fundamental challenge lies how leverage numerous images without any ground truth. To address this issue, we take an alternative approach with semi-supervised learning....
Abstract Standard registration algorithms need to be independently applied each surface register, following careful pre-processing and hand-tuning. Recently, learning-based approaches have emerged that reduce the of new scans running inference with a previously-trained model. The potential benefits are multifold: is typically orders magnitude faster than solving instance difficult optimization problem, deep learning models can made robust noise corruption, trained model may re-used for other...
The objective of this paper is to learn dense 3D shape correspondence for topology-varying generic objects in an unsupervised manner. Conventional implicit functions estimate the occupancy a point given latent code. Instead, our novel function produces probabilistic embedding represent each part space. Assuming corresponding points are similar space, we implement through inverse mapping from vector corresponded point. Both jointly learned with several effective and uncertainty-aware loss...
Learning disentangled 3D face shape representation is beneficial to attribute transfer, generation and recognition, etc. In this paper, we propose a novel distribution independence-based method learn decompose shapes. Specifically, design variational auto-encoder with Graph Convolutional Network (GCN), namely Mesh-Encoder, model the distributions of identity expression representations via inference. To disentangle facial identity, eliminate correlation two distributions, enforce them be...
Random number sequences and RNGs play an important role in trusted computing environments cryptographic applications. For example, we use random numbers the generation of keys TPM. In some web protocols, are applied to resist replay attacks. It is necessary guarantee quality their because deterministic factors likely be involved process. If a generator not designed carefully, then output may become predictable bring high security risks. Thus, design generators that produce high-quality has...
As a controllable nonlinear system, autonomous Boolean network which is easy to generate chaotic signals has become hot research topic. We realize an 18-node circuit can high entropy chaos by using the characteristics of two-input XOR gate, and establish improved mathematical model for introducing filter coefficient. Through numerical simulation comparison experiment, we study influence coefficient delay parameters on dynamic this model. The results are basically consistent with those...
Standard registration algorithms need to be independently applied each surface register, following careful pre-processing and hand-tuning. Recently, learning-based approaches have emerged that reduce the of new scans running inference with a previously-trained model. In this paper, we cast task as surface-to-surface translation problem, design model reliably capture latent geometric information directly from raw 3D face scans. We introduce Shape-My-Face (SMF), powerful encoder-decoder...