Light field imaging presents an attractive alternative to RGB because of the recording direction incoming light. The detection salient regions in a light image benefits from additional modeling angular patterns. For imaging, methods using CNNs have achieved excellent results on range tasks, including saliency detection. However, it is not trivial use CNN-based for images these are specifically designed processing inputs. In addition, current datasets sufficiently large train CNNs. To...

Camouflaged Object Detection (COD) aims to segment objects that blend in with their surroundings. Most existing methods mainly tackle this issue by a single-stage framework, which tends degrade performance the face of small objects, low-contrast and diverse appearances. In paper, we propose novel Progressive Enhancement Network (PENet) for COD imitating human visual detection system, follows three-stage process: locate refine textures restore boundary. Specifically, our PENet contains three...

This paper presents a solid-state volumetric 3-D display system utilizing 20 LC light shutters as module, which provides viewers with true depth cues. The electro-optic properties of based on normal-mode PSCT were investigated. device shows field-on transmittance 86.2%. When applied an operation voltage 100 V at 28 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">°</sup> C, the response time is 540 μs for switching-on and 180 switching-off,...

The main challenge of person re-identification (re-id) lies in the strikingly discrepancy between different camera views, including illumination, background and human pose. Existing re-id methods rely mostly on implicit solutions, such as seeking robust features or designing discriminative distance metrics. Compared to these methods, solutions are more straightforward. That is, imagine appearance target under views before matching person. key idea is that can intuitively implement viewpoint...

In this paper, a system structure and concept of smart video processing is presented to illustrated meaningful descriptions extracted from input associated with raw stream crucial the system. Then, we propose an integrated approach for moving object detection get used sensor. A statistical algorithm obtain background model, which updated real time in order adapt changes illumination objects scene. After applying threshold separate candidates foreground background, morphologic analysis area...

Conventional SPIHT compression is for a 2D image or block with the same maximum feasible decomposition level in two directions, so, only size of symmetrical, that to say directions must be and sizes had better equal 2level such as 32, 64, 128 etc, optimal performance can acquired. But, real application, images are often unsymmetrical, even have 1 line data. To overcome disadvantages conventional codec, this paper proposes modification versions codec: unsymmetrical codec 1D which adapt data...

Vehicle detection and tracking has always been a significant research on traffic surveillance video. However, multi-camera object consists of non-overlapping video network, which makes vehicle re-identification challenging problem. In this paper, we proposed novel method for continuous in campus videos. The contains two main parts: One is auto by using background modeling combining with RCNN (Region Convolutional Neural Networks). other one re-identification, collaborates visual attributes...

Scene text recognition with arbitrary shape is very challenging due to large variations in shapes, fonts, colors, backgrounds, etc. Most state-of-the-art algorithms rectify the input image into normalized image, then treat as a sequence prediction task. The bottleneck of such methods rectification, which will cause errors distortion perspective. In this paper, we find that rectification completely unnecessary. What all need spatial attention. We therefore propose simple but extremely...

Semantic issues are highly concerned with high-level interpretation in image understanding, which include text-image gap and its own affinity. Concentrating on text-formatting entities images, three sophisticated methodologies roundly reviewed as generative, discriminative descriptive grammar the basis of contextual features. The following objective benchmark for visual words is also directly presented semantic coherency. Finally, summarized directions semantics understanding discussed...

This article is about object tracking based on graph modeling. Object usually initialized by detection methods. The fundamental hypothesis that the object's pattern can be separated from its surrounding background sufficiently. However, for some objects, e.g., ball in broadcast soccer videos, it hard to extract effective features detect a single video frame. strategy adopted here identify candidate regions several consecutive frames, and then use construct relationship between regions....

The rapid increase in electric vehicles (EVs) poses significant impacts on multi-energy system (MES) operation and energy management. Accurately assessing EV charging demand becomes crucial for maintaining MES stability, making it an urgent issue to be studied. Therefore, this paper proposes a novel deep learning-based load prediction framework assess the impact of EVs MES. First, model traffic flow, modified weight fusion spatial–temporal graph convolutional network (WSTGCN) is proposed...

Code generation models have shown significant potential for programming tasks. However, existing training methods like supervised fine-tuning face key limitations: they do not effectively teach to prioritize correct over incorrect solutions in ambiguous situations, nor optimize the runtime efficiency of generated code. To address these challenges, we propose CodeDPO, a framework that integrates preference learning into code improve two factors: correctness and efficiency. CodeDPO employs...

The authors extend exemplar representation to the field of tracking and propose a robust algorithm with per‐exemplar support vector machine (SVM) classifiers. First, train simple yet effective SVM classifier using target object as single positive mining its surroundings hard negatives. Second, an online ensemble tracker, which integrates useful ‘key historical templates’ refine current template, leading better discriminative power tracker effectively decreasing risk drift. Experiments on...

In order to achieve acoustic emission detection in valve internal leakage, it is essential extract features, and establish an accurate mathematical model. Current leakage signal (VILAES) classification models mostly rely on human experience for selecting resulting low accuracy small conditions. This paper combines the wavelet scattering transform (WST), Relief-F algorithm AdaBoost.M1 algorithm, proposes use optimal coefficients as features VILAES model accurately leakage. Firstly, first...

Video saliency should be taken into consideration to facilitate optimization of the end-to-end video production, delivery and consumption ecosystem improve user experience at lowered cost. Although recent studies have significantly increased accuracy prediction, approaches are mostly video-centric, without considering any prior "bias" that viewers may with regard contents. In this paper, we propose a novel learning-based multi-modal method for optimizing user-oriented analysis. particular,...

Connecting visual imagery with descriptive language is a challenge for computer vision and machine translation. Inspired by image description, which used `encoder-decoder' model to translate into target sentence. We propose an approach that can generate descriptions video. Different from record the information in moment, video have time-serials property. So when generating we requires encoding dynamic temporal structure. Our this paper successfully takes account both global local...