- Advanced Vision and Imaging
- Image Processing Techniques and Applications
- Advanced Image Processing Techniques
- Video Surveillance and Tracking Methods
- Optical measurement and interference techniques
- Human Pose and Action Recognition
- Chaos-based Image/Signal Encryption
- Anomaly Detection Techniques and Applications
- Advanced Steganography and Watermarking Techniques
- Computer Graphics and Visualization Techniques
- Atmospheric chemistry and aerosols
- Vehicle emissions and performance
- Cell Image Analysis Techniques
- Generative Adversarial Networks and Image Synthesis
- Image and Video Stabilization
- Advanced X-ray and CT Imaging
- Digital Imaging for Blood Diseases
- Expert finding and Q&A systems
- Big Data Technologies and Applications
- Consumer Market Behavior and Pricing
- 3D Surveying and Cultural Heritage
- Medical Image Segmentation Techniques
- Sparse and Compressive Sensing Techniques
- CCD and CMOS Imaging Sensors
- Seismic Imaging and Inversion Techniques
Technical Institute of Physics and Chemistry
2025
Chinese Academy of Sciences
2025
Chengdu Medical College
2025
Chengdu University of Traditional Chinese Medicine
2025
Peking University
2024
Huazhong University of Science and Technology
2021-2024
University of British Columbia
2023
North China Institute of Science and Technology
2023
Sichuan University
2023
Beijing Forestry University
2021-2022
Flexible tactile sensors play important roles in many areas, like human-machine interface, robotic manipulation, and biomedicine. However, their flexible form factor poses challenges integration with wafer-based devices, commercial chips, or circuit boards. Here, we introduce manufacturing approaches, device designs, strategies, biomedical applications of a set flexible, modular sensors, which overcome the above achieve cooperation electronics. The exploit lithographically defined thin wires...
Fast and accurate monocular depth estimation on mobile devices is a challenging task as one should always trade off the accuracy against inference time. Most methods adopt models with large computation overhead, which are not applicable devices. However, directly training light-weight neural network to estimate can yield poor performance. To remedy this, we utilize knowledge distillation, transferring representation ability of stronger teacher student network. Experiments Mobile AI 2021...
Video depth estimation aims to infer temporally consistent depth. Some methods achieve temporal consistency by finetuning a single-image model during test time using geometry and re-projection constraints, which is inefficient not robust. An alternative approach learn how enforce from data, but this requires well-designed models sufficient video data. To address these challenges, we propose plug-and-play framework called Neural Depth Stabilizer (NVDS) that stabilizes inconsistent estimations...
As the first paper in a series on study of galaxy-galaxy lensing from Sloan Digital Sky Survey Data Release 7 (SDSS DR7), we present our image processing pipeline that corrects systematics primarily introduced by Point Spread Function (PSF). Using this pipeline, processed SDSS DR7 imaging data $r$ band and generated background galaxy catalog containing shape information each galaxy. Based own measurements images DR7, extract (GG) signals around foreground spectroscopic galaxies binned...
Current human pose estimation methods mainly use multi-scale fusion fully convolutional networks to achieve impressive results. However, this network lacks the ability capture relationship between features. In paper, we propose a method based on MLP-Mixer. detail, using 1D heatmaps as ground truth, is transformed into sequence prediction problem horizontal axis and vertical axis, so that MLP-Mixer can be directly used addition, existing backbone intra-layer fusing. order solve problem, an...
Video depth estimation aims to infer temporally consistent depth. Some methods achieve temporal consistency by finetuning a single-image model during test time using geometry and re-projection constraints, which is inefficient not robust. An alternative approach learn how enforce from data, but this requires well-designed models sufficient video data. To address these challenges, we propose plug-and-play framework called Neural Depth Stabilizer (NVDS) that stabilizes inconsistent estimations...
Microarray image processing plays an important role among array analysis and its precision will affect the application of gene chip. However, real microarray images drawn from various experimental procedures may contain noises other artifacts, which result in a lower accuracy. Hence, this paper, we propose novel denoising method by introducing compressed sensing theory into noise reduction. To improve matching pursuit efficiency, correlation coefficient is adopted. Experiments on 17...
Multi-person pose estimation is a challenging task which aims to locate keypoints for multiple persons. Graph convolutional network can effectively capture the semantic relationship among according kinematic structure of human body, beneficial but lack ability most CNN-based models. However, existing GCN-based methods mostly flatten 2D features directly obtain 1D embeddings, leading redundant information in large size and high computation cost. To address these problems, we propose two-stage...
The iteration of live streaming commerce is accelerating and the rapid development it has benefited from ever-increasing streamers joining business. Thus, we shed light on relationship between streamers’ beauty customers’ purchase behavior using hard archival data 2,597 stream
Video depth estimation aims to infer temporally consistent depth. One approach is finetune a single-image model on each video with geometry constraints, which proves inefficient and lacks robustness. An alternative learning enforce consistency from data, requires well-designed models sufficient data. To address both challenges, we introduce NVDS
The limited angle Radon transform is notoriously difficult to invert due the ill-posedness. In this work, we give a mathematical explanation that data-driven approach based on deep neural networks can reconstruct more information in stable way compared traditional methods.
Pixel art was first developed in the last century when computers have limited memory and processing power, is wide use game industry now. However, pixel a highly stylized form that requires skilled human artists to create creation process time consuming. This paper proposes an automatic way convert photos drawings reverse with unsupervised, GAN based neural networks.
Business analytics is a new term, which can be seen as an expanded filed of data science.The mathematical formulas, statistical models, and programming skills in science help companies to utilize big collect useful information from customers.However, for science, it just collects numerical values different resources, uses tools analyze the model get results.To convert numbers into information, people need use business knowledge interpret results.With clustering models customers segmented...
Being aimed at the problem that general steganography algorithms always destroy original videos after extracting information, paper adopts analysis and experiment to propose a BCH robustness method of single coefficient which can restore error bits.Before embedding data, encodes secret information with makes use prediction mode select special block control inter-frame distortion drift.Then it applies coefficients decrease modification videos.When we first make correction try recover...
At present, there is no dedicated large-scale video data set for the task of monitoring unsafe behavior in mine scenes. In this paper, miners' built by ourselves. The divided into extended group and real group. group, clear samples from 18 public sets, each with a particle simulation mask, to achieve similar dust masking effect as scene. comes site operations. This paper uses MCTN(Muti-scenic Cyclic Translation Network) complete cross-scenario knowledge transfer, so verify validity dataset...
Learning-to-Rank(LTR) is widely used in many Information Retrieval(IR) scenarios, including web search and Location Based Services(LBS) search. However, most existing LTR techniques mainly focus on homogeneous ranking. Taking QAC Dianping as an example, heterogeneous documents suggested queries (SQ) Point-of-Interests(POI) need to be ranked presented enhance user experience. New challenges are faced when conducting ranking, inconsistent feature space more serious position bias caused by...