A5102725442- Advanced Image and Video Retrieval Techniques
- Advanced MRI Techniques and Applications
- Multimodal Machine Learning Applications
- Video Surveillance and Tracking Methods
- Speech and Audio Processing
- Sparse and Compressive Sensing Techniques
- Advanced Computational Techniques and Applications
- Web Data Mining and Analysis
- Domain Adaptation and Few-Shot Learning
- Healthcare and Environmental Waste Management
- Image Processing Techniques and Applications
- Image Retrieval and Classification Techniques
- IoT-based Smart Home Systems
- Neural Networks and Applications
- EEG and Brain-Computer Interfaces
- Advanced Neural Network Applications
- Medical Imaging Techniques and Applications
- Biometric Identification and Security
- Image and Signal Denoising Methods
- Data Management and Algorithms
- Data Mining Algorithms and Applications
- Internet of Things and AI
- Advanced Image Fusion Techniques
- Embedded Systems and FPGA Design
- Speech Recognition and Synthesis
New York University
2020-2025
South Central Minzu University
2008-2022
Central China Normal University
2017
State Ethnic Affairs Commission
2017
Zhejiang University
2009
Tsinghua University
2001
Abstract Objective This study investigates speech decoding from neural signals captured by intracranial electrodes. Most prior works can only work with electrodes on a 2D grid (i.e., Electrocorticographic or ECoG array) and data single patient. We aim to design deep-learning model architecture that accommodate both surface (ECoG) depth (stereotactic EEG sEEG) The should allow training multiple participants large variability in electrode placements the trained perform well unseen during...
Abstract Objective: This study investigates speech decoding from neural signals captured by intracranial electrodes. Most prior works can only work with electrodes on a 2D grid (i.e., Electrocorticographic or ECoG array) and data single patient. We aim to design deep-learning model architecture that accommodate both surface (ECoG) depth (stereotactic EEG sEEG) The should allow training multiple participants large variability in electrode placements. not have subject-specific layers, the...
Waste auditing is important for effectively reducing the medical waste generated by resource-intensive operating rooms. To replace current time-intensive and dangerous manual method, we propose a system named iWASTE to detect classify based on videos recorded camera-equipped container. In this pilot study, collected video dataset of 4 items (gloves, hairnet, mask, shoecover) designed motion detection preprocessing method extract trim useful frames. We novel architecture R3D+C2D combining...
Building a multi-modality multi-task neural network toward accurate and robust performance is de-facto standard in perception task of autonomous driving. However, leveraging such data from multiple sensors to jointly optimize the prediction planning tasks remains largely unexplored. In this paper, we present FusionAD, best our knowledge, first unified framework that fuse information two most critical sensors, camera LiDAR, goes beyond task. Concretely, build transformer based fusion...
The reconstruction of dynamic magnetic resonance imaging (dMRI) from partially sampled k -space data has to deal with a trade-off between the spatial resolution and temporal resolution. In this paper, low-rank sparse decomposition model is introduced resolve issue, which formulated as an inverse problem regularized by robust principal component analysis (RPCA). can be solved convex optimization method. We propose scalable fast algorithm based on inexact augmented Lagrange multipliers (IALM)...
Patch-based modelling methods (i.e. sparse coding) had proven their great ability in solving the degradation problem caused by additive white Gaussian noise (AWGN) remotely sensing images. However, these usually pursue surprising performance with sacrificing computational efficiency, and learned dictionary is universal, resulting a weak model representation flexibility. To address issues, this paper proposes deep learnable (DSRD) scheme, where major difference from previous coding that...
Human robot interaction is an exciting task, which aimed to guide robots following instructions from human. Since huge gap lies between human natural language and machine codes, end models fair challenging. Further, visual information receiving sensors of also a hard for perceive. In this work, HuBo-VLM proposed tackle perception tasks associated with including object detection grounding by unified transformer based vision model. Extensive experiments on the Talk2Car benchmark demonstrate...
Multi-modality fusion is proven an effective method for 3d perception autonomous driving. However, most current multi-modality pipelines LiDAR semantic segmentation have complicated mechanisms. Point painting a quite straight forward which directly bind points with visual information. Unfortunately, previous point like methods suffer from projection error between camera and LiDAR. In our experiments, we find that this the devil in painting. As result of that, propose depth aware mechanism,...
We introduce a novel MV-DETR pipeline which is effective while efficient transformer based detection method. Given input RGBD data, we notice that there are super strong pretraining weights for RGB data less works depth related data. First and foremost , argue geometry texture cues both of vital importance could be encoded separately. Secondly, find visual feature relatively hard to extract compared with in 3d space. Unfortunately, single dataset thousands not enough training an...
Occupancy prediction tasks focus on the inference of both geometry and semantic labels for each voxel, which is an important perception mission. However, it still a segmentation task without distinguishing various instances. Further, although some existing works, such as Open-Vocabulary (OVO), have already solved problem open vocabulary detection, visual grounding in occupancy has not been to best our knowledge. To tackle above two limitations, this paper proposes Grounding (OG), novel...
Frequent Itemsets(FI) mining is a popular and important first step in analyzing datasets across broad range of applications. There are two main problems with the traditional approach for finding frequent itemsets. Firstly, it may often derive an undesirably huge set itemsets association rules. Secondly, vulnerable to noise. approaches which have been proposed address these individually. The problem addressed by Closed Itemsets(FCI), FCI removes all redundant information from result makes...
The term crease is used to denote the total line features in high resolution palmprint images. In bibliography, low images have been recognition. result limited due lack of lines poor situation, creases are abundant while they also need much more deliberate treatment. A recognition system constructed our previous research, which demonstrates feasibility using only it doesn't take into account many quality reality. this paper, we advance work image preprocessings, automatic threshold...