A5100435685- Medical Image Segmentation Techniques
- Advanced Image and Video Retrieval Techniques
- Image Retrieval and Classification Techniques
- Advanced Vision and Imaging
- Image and Signal Denoising Methods
- Video Surveillance and Tracking Methods
- Advanced Image Fusion Techniques
- Retinal Diseases and Treatments
- Retinal Imaging and Analysis
- Optical Coherence Tomography Applications
- Advanced Steganography and Watermarking Techniques
- Digital Media Forensic Detection
- Image Processing Techniques and Applications
- Glaucoma and retinal disorders
- Human Pose and Action Recognition
- Image and Video Quality Assessment
- Visual Attention and Saliency Detection
- Optical measurement and interference techniques
- Robotics and Sensor-Based Localization
- Remote-Sensing Image Classification
- Video Analysis and Summarization
- Face and Expression Recognition
- Advanced Image Processing Techniques
- Image and Object Detection Techniques
- Optical Systems and Laser Technology
Baidu (China)
2023-2025
China Southern Power Grid (China)
2024
Nanjing University of Science and Technology
2010-2023
Huazhong University of Science and Technology
2023
Guangdong University of Education
2014-2023
Guangdong University Of Finances and Economics
2014-2023
Shanghai University of Engineering Science
2022
Beijing Aerospace Flight Control Center
2019
Suzhou Industrial Park Institute of Services Outsourcing
2017
PRG S&Tech (South Korea)
2016
We address the problem of describing people based on fine-grained clothing attributes. This is an important for many practical applications, such as identifying target suspects or finding missing detailed descriptions in surveillance videos consumer photos. approach this by first mining images with attribute labels from online shopping stores. A large-scale dataset built about one million and fine-detailed sub-categories, various shades color (e.g., watermelon red, rosy purplish red), types...
In this paper, we study the problem of end-to-end multi-person pose estimation. State-of-the-art solutions adopt DETR-like framework, and mainly develop complex decoder, e.g., regarding estimation as keypoint box detection combining with human in ED-Pose [38], hierarchically predicting decoder joint (keypoint) PETR [27].We present a simple yet effective transformer approach, named Group Pose. We simply regard K-keypoint set N × K positions, each from query, well representing an instance...
The In-Context Learning (ICL) is to understand a new task via few demonstrations (aka. prompt) and predict inputs without tuning the models. While it has been widely studied in NLP, still relatively area of research computer vision. To reveal factors influencing performance visual in-context learning, this paper shows that Prompt Selection Fusion are two major have direct impact on inference learning. selection process selecting most suitable prompt for query image. This crucial because...
<para xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> This letter presents a two-stage object tracking method by combining region-based and contour-based method. First, kernel-based is adopted to locate the region. Then diffusion snake used evolve contour in order improve precision. In first localization stage, initial target position predicted evaluated Kalman filter Bhattacharyya coefficient, respectively. evolution active evolved on basis of an...
There are many problems in security of Internet Things (IOT) crying out for solutions, such as RFID tag security, wireless network transmission privacy protection, information processing security.This article is based on the existing researches technology.And it provides a new approach researchers certain IOT application and design, through analyzing summarizing ITO from various angles.
Large variations in image background may cause partial matching and normalization problems for histogram-based representations, i.e., the histograms of same category have bins which are significantly different, produce large changes differences between corresponding bins. In this paper, we deal with problem by using ratios bin values histograms, rather than values' used traditional histogram distances. We propose a ratio-based distance (BRD), is an intra-cross-bin distance, contrast previous...
The combination of pixel and superpixel has been widely utilized in the interactive segmentation methods to overcome sensitivity seeds' quantity quality. However, because introduction more variables variables' interactions, pixel-superpixel are still limited accuracy computational complexity. To solve these problems, this paper, we propose an multilabel image method. In proposed model, multilayer relationships among layer, label layer fused by Markov random field framework further improve...
This paper presents a new energy minimization method for simultaneous image segmentation and bias field estimation of magnetic resonance (MR) images. The proposed algorithm introduces the global intensity into CLIC combines local information account. target therefore is driven by two forces, one induced coherent other intensity, to ensure smoothness derived optimal improve accuracy segmentations. An integration defines an function ,then segmentations are simultaneously achieved minimizing...
To enhance the rapid assessment of geographic atrophy (GA) across macula in a single projection image generated from three-dimensional (3D) spectral-domain optical coherence tomography (SD-OCT) scans by introducing novel restricted summed-area (RSAP) technique.We describe en face GA visualization technique, RSAP, restricting axial SD-OCT images to regions beneath Bruch's membrane (BM) boundary and also considering choroidal vasculature's influence on visualization. The technique analyzes...
In this paper, we present a subcategory-aware recognition framework to boost category level object classification performance. Different from the existing monolithic model approaches, aim automatically leverage embedded subcategory structure assist further recognition. Motivated by observation of considerable intra-class diversities and inter-class ambiguities in many current data sets, explicitly split into subcategories ambiguity-guided mining. The resulting are seamlessly integrated...