A5100430822- Advanced Vision and Imaging
- Advanced Image Processing Techniques
- Handwritten Text Recognition Techniques
- Advanced Neural Network Applications
- Advanced Image and Video Retrieval Techniques
- Robotics and Sensor-Based Localization
- Image and Signal Denoising Methods
- Video Surveillance and Tracking Methods
- Generative Adversarial Networks and Image Synthesis
- Image Processing and 3D Reconstruction
- Human Pose and Action Recognition
- Optical measurement and interference techniques
- Autonomous Vehicle Technology and Safety
- Image Retrieval and Classification Techniques
- 3D Shape Modeling and Analysis
- 3D Surveying and Cultural Heritage
- Industrial Vision Systems and Defect Detection
- Vehicle License Plate Recognition
- Advanced Data Compression Techniques
- Sparse and Compressive Sensing Techniques
- Face and Expression Recognition
- Power Line Inspection Robots
- Gaze Tracking and Assistive Technology
- Grey System Theory Applications
- Energy Load and Power Forecasting
China Tobacco
2025
State Grid Corporation of China (China)
2020-2024
Shanghai Jiao Tong University
2020-2024
China University of Mining and Technology
2022-2023
Ministry of Transport
2022-2023
Kunming University of Science and Technology
2023
Amazon (Germany)
2022-2023
Huawei Technologies (United Kingdom)
2022
Shaanxi Normal University
2022
Guangdong Institute of Intelligent Manufacturing
2021
We introduce Amazon Berkeley Objects (ABO), a new large-scale dataset designed to help bridge the gap between real and virtual 3D worlds. ABO contains product catalog images, metadata, artist-created models with com-plex geometries physically-based materials that cor-respond real, household objects. derive challenging benchmarks exploit unique properties of measure current limits state-of-the-art on three open problems for real-world object understanding: single-view reconstruction, material...
This paper reviews the NTIRE 2023 challenge on image denoising (σ = 50) with a focus proposed solutions and results. The aim is to obtain network design capable produce high-quality results best performance measured by PSNR for denoising. Independent additive white Gaussian noise (AWGN) assumed level 50. had 225 registered participants, 16 teams made valid submissions. They gauge state-of-the-art
With the growing popularity of civilian unmanned aerial vehicles (UAVs), unauthorized flights are on rise accordingly. Therefore, it is critical to detect low-altitude UAVs for protecting personal privacy and public safety. Though substantial progress has been made in UAV detection, existing detection methods still have problems balancing accuracy, model size, speed. To address these limitations, this article proposes a novel deep learning method named convolution–transformer network...
Abstract As one of the hot topics in field computer vision research, face recognition technology has received significant attention due to its potentiality for a wide range applications government as well commercial purposes. In practical applications, although several existing methods have achieved good performances specific scenes, they easily suffer from sharp decline rate if affected by different conditions light, expression, posture and occlusion. Among many factors, influences complex...
The lifting scheme is well known to be an efficient tool for constructing second generation wavelets and often used design a class of biorthogonal wavelet filter banks. For its efficiency, the implementation has been adopted in international standard JPEG2000. It that orthogonality important property many applications. This paper presents how implement infinite-impulse-response (IIR) orthogonal banks by using with two steps. shown IIR can realized allpass filters Then, proposed discussed....
Traffic accidents caused by distracted driving have gradually increased in recent years. In this work, we propose a novel multi-feature fusion network based on pose estimation, for image detection. Since hand is the most important part of driver to infer actions, our proposed method firstly detects hands using human body posture information. addition features extracted from whole image, also include information and posture. The global feature, are finally fused weighted combination...
We present a graph variational autoencoder with structured prior for generating the layout of indoor 3D scenes. Given room type (e.g., living or library) and elements such as floor walls), our architecture generates collection objects furniture items sofa, table chairs) that is consistent layout. This challenging problem because generated scene needs to satisfy multiple constrains, e.g., each object should lie inside two not occupy same volume. To address these challenges, we propose deep...
In this paper, a segmentation-free keyword spotting method is proposed for Bangla handwritten documents. order to tolerate large variations in scenarios, we extracted key points based on SIFT point detector, and the end intersection found by morphological operations. Heat Kernel signature (HKS) used present local characteristics of detected points. Instead using same size patch all points, apply dynamically deciding size. Furthermore, our reduces scope searching document only considering...
Recognition of handwritten text is a useful technique that can be applied in different applications, such as signature recognition, bank check etc. However, the off-line recognition an unconstrained situation still very challenging task due to high complexity strokes and image background. This paper presents novel segmented ensembles recurrent neural network (RNN) classifiers. Two RNN models are first trained take advantage widely used geometrical feature Histogram Oriented Gradient (HOG)...
Line-structured light sensor (LLS) can provide the capability of three-dimensional point acquisition for robotics. As one type 3D scanners, LLS has been widely used in many filed robotics its strong anti-interference, fast scanning speed and high measuring accuracy. Researchers have studying methods to improve accuracy, operability years. In this paper, calibration are reviewed, which covers target plane methods. What's more, some potential improvements discussed analyzed. This review paper...
Although autonomous driving have become applicable to the industry, prevalent application of key techniques vehicles still needs be refined. For instance, how fast and accurately segment road markings in order assist next pedestrian path prediction creation high-definition (HD) map respectively is useful for more practical. Current marking segmentation mainly rely on semantic computer vision with encoder-decoder architecture. However, as demonstrated this paper, upsampling layer...
Using a layered representation for motion estimation has the advantage of being able to cope with discontinuities and occlusions. In this paper, we learn estimate optical flow by combining deep learning. Instead pre-segmenting image layers, proposed approach automatically generates using soft-mask module. The essential components module are maxout fuse operations, which enable disjoint more accurate estimation. We show that masks results in quadratic function input features output layer. can...
To get high recognition accuracy, we should train the recognizer with sufficient training data to capture characteristics of various handwriting styles and all possible occurring words. However, in most cases, available are not satisfactory enough, especially for unseen data. In this paper, try improve accuracy randomly selected data, by splitting into two parts based on trigrams recognizers separately. We also propose a modified version token passing algorithm, which makes use outputs accuracy.
Augmented reality, as an important end of merging virtual objects and the real world, has been widely used in Internet, games, e-commerce, other fields. The maturity AR technology brought a broad market space gained attention researchers companies. As ICT industry, tele-com sector also needs for its complex business scenarios issues. Unfortunately, few studies focused on application telecommunication field. This paper starts from concept technology, introduces key technologies details cases...
Recently, DNN models for lossless image coding have surpassed their traditional counterparts in compression performance, reducing the bit rate by about ten percent natural color images. But even with these advances, mathematically (MLLIC) ratios images still fall short of bandwidth and cost-effectiveness requirements most practical imaging vision systems at present beyond. To break bottleneck MLLIC we question necessity MLLIC, as almost all digital sensors inherently introduce acquisition...
Abstract Outdoor substation is an important part of power system. Substation inspection robot based on intelligent autonomous system has become the research focus unmanned inspection. In order to improve positioning accuracy and speed system, a high-precision algorithm transformer detection proposed in this paper. Tikhonov regularization used correct pathological problem localization model. The observation amount receiver increased by using four signals single base station with double...
Object detection plays an important role in underground intelligent vehicles and transportation systems. Due to the uneven light mining scenarios, infrared cameras are one of typical onboard sensors for environmental perception. Although object has been studied decades, it still confronts challenge detecting objects mines. The contributing factors include weak small images similar environments scenarios. In this paper, a Feature Enhancement Guided Network (FEGNet) is proposed address these...
In the realm of intelligent vehicles, evolution object detection algorithms is paramount importance. Current deep learning-based methodologies excel in identifying medium to large-sized objects but often falter with smaller entities. A notable research gap exists integrating key vehicular state data, such as velocity and steering angle, into generally designed frameworks. To bridge these gaps, we present Prior-YOLO, a novel modification YOLO v8, marked by advanced network structure refined...