A5010965630- Crystallization and Solubility Studies
- X-ray Diffraction in Crystallography
- Speech and Audio Processing
- Complex Network Analysis Techniques
- Advanced Vision and Imaging
- Image Processing Techniques and Applications
- Advanced Neural Network Applications
- Cell Image Analysis Techniques
- Opinion Dynamics and Social Influence
- Direction-of-Arrival Estimation Techniques
- Face recognition and analysis
- Advanced Image and Video Retrieval Techniques
- Recommender Systems and Techniques
- Face and Expression Recognition
- Sparse and Compressive Sensing Techniques
- Indoor and Outdoor Localization Technologies
- Robotics and Sensor-Based Localization
- Mobile Ad Hoc Networks
- Advanced Clustering Algorithms Research
- Digital Imaging for Blood Diseases
- Blind Source Separation Techniques
- Image Retrieval and Classification Techniques
- Advanced Fluorescence Microscopy Techniques
- Multimodal Machine Learning Applications
- Additive Manufacturing Materials and Processes
Amazon (United States)
2022-2025
Northeastern University
2012-2024
GE Global Research (United States)
2021-2024
Beijing University of Posts and Telecommunications
2024
Fudan University
2021-2023
Linyi University
2023
University of Toronto
2022
University of Nevada, Las Vegas
2018-2022
Xidian University
2014-2021
Johns Hopkins University
2021
Fusing LiDAR and camera information is essential for accurate reliable 3D object detection in autonomous driving systems. This challenging due to the difficulty of combining multi-granularity geometric semantic features from two drastically different modalities. Recent approaches aim at exploring densities through lifting points 2D images (referred as "seeds") into space, then incorporate semantics via cross-modal interaction or fusion techniques. However, depth under-investigated these when...
In an adaptive beamforming system, the mutual coupling effects among array elements can seriously degrade system performance. this paper, we propose a robust algorithm using uniform linear (ULA) to mitigate effects. The proposed is based on fact that matrix (MCM) of ULA be approximated as banded symmetric Toeplitz between two sensors inversely related their separation and negligible for few wavelengths away. By exploiting structural characteristics MCM, subspace-based method used estimate...
Abstract Motivation While multi-channel fluorescence microscopy is a vital imaging method in biological studies, the number of channels that can be imaged simultaneously limited by technical and hardware limitations such as emission spectra cross-talk. One solution using deep neural networks to model localization relationship between two proteins so one protein digitally predicted. Furthermore, input predicted implicitly reflect modeled relationship. Accordingly, observing response...
The clustering phenomenon is common in real world networks. A discrete-time network model proposed firstly this paper, and then the phase dynamics of networks are studied carefully. acts as a bridge between dynamic topology modular network. On one hand, will occur for by model; on other communities can be identified from phenomenon. Beyond phases' information, it found that frequencies phases applied to community detection also with model. In specific, information their nodes. Detailed...
This paper addresses the challenging unsupervised scene flow estimation problem by jointly learning four low-level vision sub-tasks: optical F, stereo-depth D, camera pose P and motion segmentation S. Our key insight is that rigidity of shares same inherent geometrical structure with object movements depth. Hence, from S can be inferred coupling D to achieve more robust estimation. To this end, we propose a novel framework named EffiScene efficient joint learning, going beyond existing...
Referring Image Segmentation (RIS) aims at segmenting the target object from an image referred by one given natural language expression. The diverse and flexible expressions complex visual contents in images raise RIS model with higher demands for investigating fine-grained matching behaviors between words objects presented images. However, such are hard to be learned captured when cues of referents (i.e. objects) insufficient, as weak tend easily confused cluttered background boundary or...
Anode materials have a vital influence on the performance of sodium ion batteries. In this paper, SnSb nanoparticles were distributed uniformly in N-doped three-dimensional porous carbon (SnSb@N-PC), which effectively avoided agglomeration alloy and greatly improved capacity retention rate SnSb@N-PC. At same time, substrate brings higher conductivity, larger specific surface area, more storage sites, makes material obtain excellent properties. The first discharge-specific SnSb@N-PC was 846.3...
Single Shot MultiBox Detector (SSD) is an effective method for multi-targets detection. However, due to the small scale and less information, targets detection a difficult assignment SSD. In order address this issue, we design framework based on SSD, namely feature fusion enhancement First, considering that information contained in different maps different, especially shallow which contains more detailed lacks semantic propose of fusing introduce abundant object information. Secondly,...
A novel least square (LS) localization method for wireless sensor networks (WSNs) using received signal strength indicator (RSSI) is proposed in this paper. Unlike previous LS methods, the performs location calculation with aid of condition number coordinate matrix to avoid appearance outliers. The threshold introduced paper outliers while enhance accuracy. Simulation results demonstrate that our approach can suppress more efficiently, and improve accuracy stability localization.
In this paper, a novel method for two-dimensional (2-D) direction-of-arrival (DOA) estimation employing sparse L-shaped array is proposed. The composed of one traditional uniform linear (ULA), ULA and an auxiliary sensor. the proposed method, elevation angles are estimated by two steps. first step to estimate ambiguous angle ESPRIT-like method. second remove ambiguity disambiguation procedure. Then, azimuth estimates all data cross-covariance matrix (CCM) without any pairing procedures....
Discriminative facial parts are essential for expression recognition (FER) tasks because of small inter-class differences and large intra-class variations in images. Existing methods localize discriminative regions with the aid extra landmarks, such as action units (AU). However, it consumes a lot manpower manually labeling. To address this problem, paper, we propose an advanced attention based convolutional neural network (FA-CNN) 2D+3D FER. The main contribution FA-CNN is mechanism, which...