A5080616233- Human Pose and Action Recognition
- Video Surveillance and Tracking Methods
- Advanced Neural Network Applications
- Advanced Image and Video Retrieval Techniques
- Image Retrieval and Classification Techniques
- Anomaly Detection Techniques and Applications
- Advanced Vision and Imaging
- Gait Recognition and Analysis
- Face recognition and analysis
- Image and Object Detection Techniques
- Domain Adaptation and Few-Shot Learning
- Advanced Text Analysis Techniques
- Visual Attention and Saliency Detection
- Multimodal Machine Learning Applications
- Web Data Mining and Analysis
- Fungal and yeast genetics research
- Monoclonal and Polyclonal Antibodies Research
- Biochemical and Molecular Research
- Bacterial Genetics and Biotechnology
- Natural Language Processing Techniques
- 3D Shape Modeling and Analysis
- Glycosylation and Glycoproteins Research
- Industrial Vision Systems and Defect Detection
- Vehicle License Plate Recognition
- HER2/EGFR in Cancer Research
Jingdong (China)
2023
Silicon Valley Community Foundation
2019-2021
University of South Carolina
2014-2019
Lanzhou University
2017
Human visual perception shows good consistency for many multi-label image classification tasks under certain spatial transforms, such as scaling, rotation, flipping and translation. This has motivated the data augmentation strategy widely used in CNN classifier training -- transformed images are included by assuming same class labels their original images. In this paper, we further propose assumption of perceptual attention regions i.e., region a follows transform if input is spatially...
We propose a new learning-based method for estimating 2D human pose from single image, using Dual-Source Deep Convolutional Neural Networks (DS-CNN). Recently, many methods have been developed to estimate by priors that are estimated physiologically inspired graphical models or learned holistic perspective. In this paper, we integrate both the local (body) part appearance and view of each more accurate estimation. Specifically, proposed DS-CNN takes set image patches (category-independent...
We propose a new learning-based method for estimating 2D human pose from single image, using Dual-Source Deep Convolutional Neural Networks (DS-CNN). Recently, many methods have been developed to estimate by priors that are estimated physiologically inspired graphical models or learned holistic perspective. In this paper, we integrate both the local (body) part appearance and view of each more accurate estimation. Specifically, proposed DS-CNN takes set image patches (category-independent...
In the past decade, Convolutional Neural Networks (CNNs) have been demonstrated successful for object detections. However, size of network input is limited by amount memory available on GPUs. Moreover, performance degrades when detecting small objects. To alleviate usage and improve traffic signs, we proposed an approach signs from large images under real world conditions. particular, are broken into patches as to a Small-Object-Sensitive-CNN (SOS-CNN) modified Single Shot Multibox Detector...
MBS301, a glyco-engineered bispecific anti-human epidermal growth factor receptor 2 (HER2) antibody with typical IgG1 monoclonal structure, was developed through dual-cell expression and in vitro assembling process. MBS301 consists of two half antibodies engineered from trastuzumab pertuzumab, respectively. Integrity purity profiles MB301 indicated that the heterodimerization successful. The high similar melting temperatures (Tm1,72.0°C Tm2, 84.8°C) compared those its parental pertuzumab...
Automatic tracking of large-scale crowded targets are particular importance in many applications, such as people/vehicle video surveillance, fiber materials science, and cell biomedical imaging. This problem becomes very challenging when the show similar appearance interslice/ inter-frame continuity is low due to sparse sampling, camera motion target occlusion. The main challenge comes from step association which aims at matching predictions observations multiple targets. In this paper we...
The dissacharide trehalose is an important intracellular osmoprotectant and the OtsA/B pathway principal for biosynthesis in a wide range of bacterial species. Scaffolding proteins other cytoskeletal elements play essential role morphogenetic processes bacteria. Here we describe how OtsA, addition to its biosynthesis, functions as osmotic stress sensor regulate cell morphology Arthrobacter strain A3. In response stress, this species undergo transition from bacillary myceloid growth. An otsA...
In this paper, we study the problem of Cross-View Person Identification (CVPI), which aims at identifying same person from temporally synchronized videos taken by different wearable cameras. Our basic idea is to utilize human motion consistency for CVPI, where can be computed optical flow. However, flow view-variant - person's in very due view angle change. attempt 3D human-skeleton sequences learn a model that extract view-invariant features flows views. For purpose, use Mocap database...
Wearable cameras, such as Google Glass and Go Pro, enable video data collection over larger areas from different views. In this paper, we tackle a new problem of locating the co-interest person (CIP), i.e., one who draws attention most camera wearers, temporally synchronized videos taken by multiple wearable cameras. Our basic idea is to exploit motion patterns people use them correlate persons across videos, instead performing appearance-based matching in traditional...
We propose a novel domain-specific generative pre-training (DS-GPT) method for text generation and apply it to the product titleand review summarization problems on E-commerce mobile display.First, we adopt decoder-only transformer architecture, which fitswell fine-tuning tasks by combining input output all to-gether. Second, demonstrate utilizing only small amount of data in related domains is powerful. Pre-training languagemodel from general corpus such as Wikipedia or CommonCrawl requires...
Video-based human action recognition benefits from multiple cameras which can provide temporally synchronized, multi-view videos. Cross-video person identification, i.e., determining whether at a given time, persons tracked in different videos are the same or not, is key step to integrate such information for collaborative recognition. For fixed cameras, this relatively easy since be precalibrated. In paper, we study cross-video identification wearable constantly moving with wearers....
In the past decade, Convolutional Neural Networks (CNNs) have been demonstrated successful for object detections. However, size of network input is limited by amount memory available on GPUs. Moreover, performance degrades when detecting small objects. To alleviate usage and improve traffic signs, we proposed an approach signs from large images under real world conditions. particular, are broken into patches as to a Small-Object-Sensitive-CNN (SOS-CNN) modified Single Shot Multibox Detector...
In this paper, we proposed an automatic Scenario-based Multi-product Advertising Copywriting Generation system (SMPACG) for E-Commerce, which has been deployed on a leading Chinese e-commerce platform. The SMPACG consists of two main components: 1) multi-product combination selection module, itself is consisted topic prediction model, pattern and attribute-based model arbitrator model; 2) advertising copywriting generation combines our domain-specific pretrained language knowledge-based data...
Product images are essential for providing desirable user experience in an e-commerce platform. For a platform with billions of products, it is extremely time-costly and labor-expensive to manually pick organize qualified images. Furthermore, there the numerous complicated image rules that product needs comply order be generated/selected. To address these challenges, this paper, we present new learning framework achieve Automatic Generation Product-Image Sequence (AGPIS) e-commerce. end,...
Region-based Convolutional Neural Networks (R-CNNs) have achieved great success in the field of object detection. The existing R-CNNs usually divide a Region-of-Interest (ROI) into grids, and then localize objects by utilizing spatial information reflected relative position each grid ROI. In this paper, we propose novel feature-encoding approach, where is represented through distributions visual patterns. particular, design Mask Weight Network (MWN) to learn set masks apply channel-wise...
Wearable cameras, such as Google Glass and Go Pro, enable video data collection over larger areas from different views. In this paper, we tackle a new problem of locating the co-interest person (CIP), i.e., one who draws attention most camera wearers, temporally synchronized videos taken by multiple wearable cameras. Our basic idea is to exploit motion patterns people use them correlate persons across videos, instead performing appearance-based matching in traditional...