An auto fabric defect detection system via computer vision is used to replace manual inspection. In this paper, we propose a hardware accelerated algorithm based on small-scale over-completed dictionary (SSOCD) sparse coding (SC) method, which realized parallel platform (TMS320C6678). order reduce computation, the image patches projections in training SSOCD are taken as features and proposed more robust, exhibit obvious advantages results computational cost. Furthermore, introduce ratio...

To overcome the interference caused by varying lighting conditions in human pose estimation (HPE), significant advancements have been made event-based approaches. However, since event cameras are only sensitive to illumination changes, static bodies often lead motion ambiguity, making it challenging for existing methods handle such cases effectively. Therefore, we propose EvTransPose, a novel framework that combines an hourglass module global dependencies and pyramid encoding local features....

This paper proposed a defect detection algorithm for fabrics with complex texture based on dual-scale over-complete dictionary. The core was to learn the features of defect-free using traditional methods generally showed favorable effect plain cloth or twill, etc., while poor effects plaids stripes, etc. texture. Considering large variations size different kinds, this study used dictionary enhance self-adaptability detection. Subsequently, increase in false rate brought by effectively...

3D object detection in LiDAR point clouds has been extensively used autonomous driving, intelligent robotics, and augmented reality. Although the one-stage detector satisfactory training inference speed, there are still some performance problems due to insufficient utilization of bird’s eye view (BEV) information. In this paper, a new backbone network is proposed complete cross-layer fusion multi-scale BEV feature maps, which makes full use various information for detection. Specifically,...

Automated segmentation and evaluation of carotid plaques ultrasound images is great significance for the diagnosis early intervention high-risk groups cardiovascular cerebrovascular diseases. However, it remains challenging to develop such solutions due relatively low quality heterogenous characteristics plaques. To address those problems, in this paper, we propose a novel deep convolutional neural network, FRDD-Net, with an encoder-decoder architecture automatically segment We feature...

Intelligent tutoring systems (ITS) provide individualized instruction. They offer many advantages over the traditional classroom scenario: they are always available, nonjudgmental and tailored feedback resulting in increased effective learning. However, still not as one-on-one human tutoring. The next generation of intelligent tutors is expected to be able take into account cognitive emotional state students. We present a proposed contribution affect student modeling, reports on progress...

Automatic aircraft engine defect detection is a challenging but important task in industry which can ensure safe air transportation and flight. In this paper, we propose fast accurate feature weighting network (FWNet) to solve the problem of scale variation improve accuracy. The framework designed based on recent popular convolutional neural networks pyramid. To further boost representation power network, new module (FWM) was proposed recalibrate channel-wise attention increase weights valid...

Automatic ship detection in optical remote sensing images is of great significance due to its broad applications maritime security and fishery control. Most algorithms utilize a single-band image design low-level hand-crafted features, which are easily influenced by interference like clouds strong waves not robust for large-scale variation ships. In this paper, we propose novel coarse-to-fine method based on discrete wavelet transform (DWT) deep residual dense network (DRDN) address these...

High dynamic range imaging (HDRI) is an essential task in remote sensing, enhancing low (LDR) sensing images and benefiting downstream tasks, such as object detection image segmentation. However, conventional frame-based HDRI methods may encounter challenges real-world scenarios due to the limited information inherent a single captured by cameras. In this paper, event-based HDR framework proposed address problem, denoted ERS-HDRI, which reconstructs from single-exposure LDR its concurrent...

Image deblurring is a challenging problem in image processing, which aims to reconstruct an original high-quality from its blurred measurement caused by various factors, for example, imperfect focusing the imaging system or different depths of scene appearing commonly our daily photos. Recently, sparse representation whose basic idea code patch as linear combination few atoms chosen out over-complete dictionary has shown uplifting results deblurring. Based on this and another heart-stirring...

Spatiotemporal graph convolution has made significant progress in skeleton-based action recognition recent years. Most of the existing methods take all joints human skeleton as overall modeling graph, ignoring differences movement patterns various parts human, and cannot well connect relationship between different skeleton. To capture unique features data correlation parts, we propose two new methods: whole network (WGCN) part (PGCN). WGCN learns scale spatiotemporal according to physical...

Deep convolutional neural networks have recently achieved great success on image aesthetics assessment task. In this paper, we propose an efficient method which takes the global, local and scene-aware information of images into consideration exploits composite features extracted from corresponding pretrained deep learning models to classify derived with support vector machine. Contrary popular methods that require fine-tuning or training a new model scratch, our training-free directly...

Within the development of deep convolutional neural network, great achievements had been made in single-image super-resolution (SISR) task. However, higher performance always comes with deeper layers which also brings larger numbers network operations and parameters that make it hard to implement practice. In our work, a lightly super-resolution, named Mobile Share- Source Network (MSSN), is purposed address these practical issues. MSSN, high-efficiency block, mobile adaptive weighted...

In this work, we present a convolutional neural network (CNN) named CGFA-CNN for blind image quality assessment (BIQA). A unique two-stage strategy is utilized which firstly identifies the distortion type in an using Sub-Network I and then quantifies II. Different from most deep networks, extract hierarchical features as descriptors to enhance representation design feature aggregation layer end-to-end training manner applying Fisher encoding visual vocabularies modeled by Gaussian mixture...

Deep neural networks are vulnerable to the adversarial example synthesized by adding imperceptible perturbations original image but can fool classifier provide wrong prediction outputs. This paper proposes an restoration approach which provides a strong defense mechanism robustness against attacks. We show that unsupervised framework, deep prior, effectively eliminate influence of perturbations. The proposed method uses multiple prior called tandem priors recover from example. Tandem contain...

Three-dimensional object detection from point cloud data is becoming more and significant, especially for autonomous driving applications. However, it difficult lidar to obtain the complete structure of an in a real scene due its scanning characteristics. Although existing methods have made great progress, most them ignore prior information structure, such as symmetry. So, this paper, we use symmetry missing part then detect it. Specifically, propose two-stage framework. In first stage,...