A5100372739- Advanced Image and Video Retrieval Techniques
- Video Analysis and Summarization
- Image Retrieval and Classification Techniques
- Video Surveillance and Tracking Methods
- Particle Detector Development and Performance
- Multimodal Machine Learning Applications
- Dark Matter and Cosmic Phenomena
- Advanced Neural Network Applications
- Radiation Detection and Scintillator Technologies
- Nuclear Physics and Applications
- Human Pose and Action Recognition
- Face and Expression Recognition
- Algorithms and Data Compression
- Face recognition and analysis
- Digital Media Forensic Detection
- Domain Adaptation and Few-Shot Learning
- CCD and CMOS Imaging Sensors
- Music and Audio Processing
- Generative Adversarial Networks and Image Synthesis
- Atomic and Subatomic Physics Research
- Advancements in PLL and VCO Technologies
- Particle physics theoretical and experimental studies
- Neutrino Physics Research
- Speech Recognition and Synthesis
- Nuclear physics research studies
Tsinghua University
2016-2025
China Academy of Building Research
2023-2025
Dalian University of Technology
2023
Beijing Academy of Artificial Intelligence
2022
Robert Bosch (United States)
2021
Ministry of Education of the People's Republic of China
2020
Qinghai Normal University
2019
Intelligent Health (United Kingdom)
2016
Shandong Jianzhu University
2012
Tianjin University
2011
This paper conducts a formal study of the shot boundary detection problem. First, general framework techniques is proposed. Three critical techniques, i.e., representation visual content, construction continuity signal and classification values, are identified formulated in perspective pattern recognition. Meanwhile, major challenges to identified. Second, comprehensive review existing approaches conducted. The representative categorized compared according their roles framework. Based on...
The China Jinping Underground Laboratory, inaugurated in 2010, is an underground research facility with the deepest rock overburden and largest space by volume world. first-generation science programs include dark matter searches conducted CDEX PandaX experiments. These activities are complemented measurements of ambient radioactivity installation low-background counting systems. Phase II being constructed, its potential projects formulated. In this review, we discuss history, key features,...
The two-stage methods for instance segmentation, e.g. Mask R-CNN, have achieved excellent performance recently. However, the segmented masks are still very coarse due to downsampling operations in both feature pyramid and instance-wise pooling process, especially large objects. In this work, we propose a new method called RefineMask high-quality segmentation of objects scenes, which incorporates fine-grained features during segmenting process multi-stage manner. Through fusing more detailed...
Tremendous efforts have been made on instance segmentation but the mask quality is still not satisfactory. The boundaries of predicted masks are usually imprecise due to low spatial resolution feature maps and imbalance problem caused by extremely proportion boundary pixels. To address these issues, we propose a conceptually simple yet effective post-processing refinement framework improve based results any model, termed BPR. Following idea looking closer segment better, extract refine...
Thermal infrared detection systems play an important role in many areas such as night security, autonomous driving, and body temperature detection. They have the unique advantages of passive imaging, sensitivity penetration. But security these themselves has not been fully explored, which poses risks applying systems. We propose a physical attack method with small bulbs on board against state of-the-art pedestrian detectors. Our goal is to make detectors unable detect real-world pedestrians....
Thermal infrared imaging is widely used in body temperature measurement, security monitoring, and so on, but its safety research attracted attention only recent years. We proposed the adversarial clothing, which could fool pedestrian detectors at different angles. simulated process from cloth to clothing digital world then designed "QR code" pattern. The core of our method design a basic pattern that can be expanded periodically, make after random cropping deformation still have an effect,...
We propose an approach for traffic sign detection based on Convolutional Neural Networks (CNN). first transform the original image into gray scale by using support vector machines, then use convolutional neural networks with fixed and learnable layers recognition. The layer can reduce amount of interest areas to detect, crop boundaries very close borders signs. increase accuracy significantly. Besides, we bootstrap methods improve avoid overfitting problem. In German Traffic Sign Detection...
We present a traffic sign detection model consisting of two modules. The first module is for ROI (region interest) extraction. By supervised learning, it transforms the color images to gray such that characteristic colors signs are more distinguishable in images. It follows shape template matching, where set templates each target category designed. After that, ROIs generated. second recognition. validates if an belongs by learning. Local and features extracted. learning methods used SVMs....
In the field of construction engineering, there exists a dynamic evolution extensive technical standards and specifications (e.g., GB/T ISO series) that permeate entire lifecycle design, construction, operation–maintenance. These require continuous version iteration to adapt technological innovations. Engineers specialized knowledge bases assist in understanding updating these standards. The advancement large language models (LLMs) Retrieval-Augmented Generation (RAG) technologies provides...
About ten years ago, HMAX was proposed as a simple and biologically feasible model for object recognition, based on how the visual cortex processes information. However, does not encompass sparse firing, which is hallmark of neurons at all stages pathway. The current paper presents an improved model, called HMAX, integrates firing. This able to learn higher-level features objects unlabeled training images. Unlike most other deep learning models that explicitly address global structure images...
During 2013-2015 an expansion of the China JinPing underground Laboratory (CJPL) will be undertaken along a main branch bypass tunnel in complex. This second phase CJPL increase laboratory space to approximately 96,000 m3, which can compared existing CJPL-I volume ∼ 4,000 m3. One design configuration has eight additional hall spaces, each over 60 m long and approximately12 width, with overburdens about 2.4 km rock, oriented parallel away from water transport auto traffic tunnels. There are...
Person re-identification is a hot topic in computer vision, and the loss function plays vital role improving discrimination of learned features. However, most existing models utilize hand-crafted functions, which are usually sub-optimal challenging to be designed. In this paper, we propose novel method, AutoLoss-GMS, search better space generalized margin-based softmax for person reidentification automatically. Specifically, first decomposed into two computational graphs constant. Then...
The direct detection of light dark matter is a prominent topic in international physics research. Weakly Interacting Massive Particles (WIMPs) are the leading candidates. scientific goal China Dark EXperiment (CDEX), located Jinping Underground Laboratory (CJPL), to detect candidates using high-purity germanium detector array. In 2013, CDEX-1B achieved most sensitive experimental results for spin-independent WIMPs below 4 GeV/<italic xmlns:mml="http://www.w3.org/1998/Math/MathML"...
In this paper we conduct a relatively complete study on how to exploit spatial context constraints for automated image region annotation. We present straight forward method regularize the segmented regions into 2D lattice layout, so that simple grid-structure graphical models can be employed characterize dependencies. show represent in various and also related learning inference algorithms. Different from most of existing work, specifically investigate combine classification performance...
Though both quantity and quality of semantic concept detection in video are continuously improving, it still remains unclear how to exploit these detected concepts as indices search, given a specific query. In this paper, we tackle problem propose search framework which operates like searching text documents. Noteworthy for its adoption the well-founded principles, first selects few related query, by employing tf-idf scheme, called c-tf-idf, measure informativeness These selected form...
Min-wise hash is a widely-used hashing method for scalable similarity search in terms of Jaccard similarity, while practice it necessary to compute many such functions certain precision, leading expensive computational cost. In this paper, we introduce an effective method, i.e. the min-max which significantly reduces time by half, yet has provably slightly smaller variance estimating pair wise similarity. addition, estimator only contains equality checking, thus especially suitable...
This paper presents an automatic temperature correction design for carry chain based time-to-digital converter (TDC) in field-programmable gate array (FPGA). The bin-by-bin calibrations under different temperatures are performed both plain TDC and Wave Union to characterize the influence of variation on delay time chain. Accordingly, a simplified scheme by using dedicated channel measure coefficient correct fine result all channels is implemented tested. method shows only few picosecond...
Drogue detection is a fundamental issue during the close docking phase of autonomous aerial refueling (AAR). To cope with this issue, novel and effective method based on deep learning convolutional neural networks (CNNs) proposed. In order to ensure its robustness wide application, dataset images was prepared by utilizing real data "Probe Drogue" refueling, which contains diverse drogues in various environmental conditions without artificial features placed drogues. By employing ideas...