Prevailing video frame interpolation algorithms, that generate the intermediate frames from consecutive inputs, typically rely on complex model architectures with heavy parameters or large delay, hindering them diverse real-time applications. In this work, we devise an efficient encoder-decoder based network, termed IFRNet, for fast in-termediate synthesizing. It first extracts pyramid features given and then refines bilateral flow fields together a powerful intermedi-ate feature until...

Global contrast considers the color difference between a target region or pixel and rest of image. It is frequently used to measure saliency pixel. In previous global contrast-based methods, usually measured by sum from entire We find that spatial distribution one important cue neglected works. Foreground has high all directions, since it surrounded background. Background often shows low in at least direction, as connect Motivated this intuition, we first compute directional different...

The brain signal classification is the basis for implementation of brain-computer interfaces (BCIs). However, most existing methods are based on processing technology, which require a significant amount manual intervention, such as channel selection and dimensionality reduction, often struggle to achieve satisfactory accuracy. To high accuracy little intervention possible, convolutional dynamically convergent differential neural network (ConvDCDNN) proposed solving electroencephalography...

<abstract><p>Salient object detection (SOD) aims to detect the most attractive region in an image. Fully supervised SOD based on deep learning usually needs a large amount of data with human annotation. Researchers have gradually focused task using weakly annotation such as category, scribble, and bounding-box, while these existing methods achieve limited performance demonstrate huge gap fully methods. In this work, we proposed one novel two-stage method bounding-box recent...

Since deformation estimation may lead to errors occurring when the camera vibrates, it is necessary remove image global motion before computing real bridge deformation. In this study, a combination of correction algorithm and 2D image-based measurement technique was utilized address issue during data acquisition for measurement. Based on proposed methodology, parameters were estimated by defining an effective sub-image in using Iterative Affine Motion Estimator. Then applied all pixels each...

Along with current multi-scale based detectors, Feature Aggregation and Enhancement (FAE) modules have shown superior performance gains for cutting-edge object detection. However, these hand-crafted FAE show inconsistent improvements on face detection, which is mainly due to the significant distribution difference between its training applying corpus, i.e. COCO vs. WIDER Face. To tackle this problem, we essentially analyse effect of data distribution, consequently propose search an effective...

Text detection and recognition in natural scene images plays an important role content analysis of images. In this paper, based on the characteristics text, we propose a robust text method using Maximally Stable Extremal Regions (MSER) Support Vector Machine (SVM). Different from end to recognition, split problem into procedure. Firstly, stage, order extract potential as much possible, use MSER color clustering connected component. Then, for obtained candidate component, visual saliency some...

Enhancing the resolution of images by super-resolution reconstruction algorithm is less costly compared with way upgrading hardware devices. At present, image can recover texture details well, but performance relatively general on remote sensing lower contrast and more complex texture, reconstructed are prone to noise, missing checkerboard effect. This paper proposes a adversarial network based inverse feature fusion for characteristics images, which combines high-level semantics low-level...

Motion information is one important cue in unsupervised video salient object detection. In order to estimate motion videos, most of the methods adopt time-consuming algorithms such as large displacement optical flow estimation (needs more than 8-40s with 640X480 size per frame), which leads saliency detection only 0.01-0.1 FPS speed and limits its application. human visual system, usually considered a whole. Therefore, we need not compute each pixel. Instead, it desirable probability pixel...

The human visual system tends to consider saliency of an object as a whole. Some object-level detection methods have been proposed by leveraging proposals in bounding boxes, and regarding the entire box one candidate salient region. However, boxes can not provide exact position lot pixels belong background. Consequently, background also show high saliency. Besides, acquiring needs time cost. In order compute saliency, we region growing from some seed superpixels, find surrounding which...

Table detection is of importance in the field document images analysis and processing, especially table frame line detection. Although a great success has been achieved for high quality during past decade, low still remains challenge. To address this problem, we proposed neoteric method to detect automatically images. Firstly, Radon transform adopted skew then correct it. Secondly, run length smoothing algorithm (RLSA) used extract lines longer than predefined threshold. Thirdly, locate...

This paper presents a novel Physically-guided Disentangled Implicit Rendering (PhyDIR) framework for highfidelity 3D face modeling. The motivation comes from two observations: Widely-used graphics renderers yield excessive approximations against photo-realistic imaging, while neural rendering methods produce superior appearances but are highly entangled to perceive 3D-aware operations. Hence, we learn disentangle the implicit via explicit physical guidance, guaranteeing properties of: (1)...

Although a great success has been achieved for the situation of high quality images during past decades, Character recognition in low still remains challenge. To tackle this challenge, paper novel method SVM framework is proposed to recognize characters document by using local and global features. Firstly, multi-scale sliding window strategy with pruning character traits adopted generate potential sub-regions. Then, conventional feature state-of-art feature, namely histogram oriented...

This paper presents a fast non-local disparity refinement method based on belief propagation. The propagated minimum spanning tree only need two sequential passes, first from leaf nodes to root, then root nodes. Computational complexity of each pixel at all levels is O(1). Performance evaluation standard Middlebury data sets shows that the proposed outperforms local both in accuracy and speed. Compared with existing nonlocal method, about maximum 15× faster speed almost same accuracy.

<abstract><p>The rapid development of deep learning has made a great progress in salient object detection task. Fully supervised methods need large number pixel-level annotations. To avoid laborious and consuming annotation, weakly consider low-cost annotations such as category, bounding-box, scribble, etc. Due to simple annotation existing large-scale classification datasets, the category based have received more attention while still suffering from inaccurate detection. In this...

Abstract Amidst the escalating demand for energy and proliferation of renewable sources, direct current (DC) systems have garnered increasing interest. As pivotal safety reliability DC systems, testing circuit breakers is a critical step to evaluate their performance dependability. This study focuses on breaking test method medium high-voltage breakers. It compares evaluates waveform, recovery voltage, dissipation across various methodologies during process assesses suitability methods...

The YOLOv3 algorithm is widely used in the industry due to its high speed and precision. Aiming at problem of low detection accuracy slow rate wearing helmets intelligent monitoring, a YOLOv3N based on improved (You Only Look Once) proposed. Improve network structure basis algorithm, replace Darknet-53 traditional convolution with fewer parameters, reduce model increase rate; order screen out required frames more reasonably, NMS optimized. Experimental results show that compared YOLOv3,...

Object proposals generation plays an important role in computer vision. A good object model should assign obviously high and low objectness score to the window that contains complete objects incomplete objects, respectively. However, some existing methods such as local contrast based models usually fail satisfy this requirement. In letter, we propose MBDSal Box, a minimum barrier distance (MBD) saliency box for locating proposals. Box consists of three components: item: First, computation...

This paper presents a novel full-image guided filtering based on eight-connected weight propagation for dense stereo matching. The proposed method has three main features: first, the is more approximate compared to previous approach, second, pixels employed into are all without constrained by one fixed window, last but not least, computational complexity of each pixel at disparity level 0(1), and implementation filter can efficiently parallelized hardware platform. Performance evaluation...

To ascertain the cognition pattern of novice driver, indoor tests were designed in this study. Subjects' reaction time during ongoing test was recorded. By using H6 Head Mounted Optics eye tracking system, subjects' movement data recorded; at same time, EEG recorded by BioGraph Infiniti physical feedback system. With field data, paper analyzed correlation between average and strokes Chinese character (Ns), total number information on sign (Nt) both placename searching placenames reading...