In this paper, we present a Character-Aware Neural Network (Char-Net) for recognizing distorted scene text. Our Char-Net is composed of word-level encoder, character-level and LSTM-based decoder. Unlike previous work which employed global spatial transformer network to rectify the entire text image, take an approach detecting rectifying individual characters. To end, introduce novel hierarchical attention mechanism (HAM) consists recurrent RoIWarp layer layer. The sequentially extracts...

One of the most challenging issues in color guided depth map restoration is inconsistency between edges guidance images and discontinuities on maps. This makes restored suffer from texture copy artifacts blurring discontinuities. To handle this problem, state-of-the-art methods design complex weight based heuristically make use bicubic interpolation input map. In paper, we show that using interpolated can blur when upsampling factor large contains holes heavy noise. contrast, propose a...

Edge-preserving image smoothing is a fundamental procedure for many computer vision and graphic applications. There tradeoff between the quality processing speed: high usually requires computational cost, which leads to low speed. In this article, we propose new global optimization based method, named iterative least squares (ILS), efficient edge-preserving smoothing. Our approach can produce high-quality results but at much lower cost. Comprehensive experiments demonstrate that proposed...

Image smoothing is a fundamental procedure in applications of both computer vision and graphics. The required properties can be different or even contradictive among tasks. Nevertheless, the inherent nature one operator usually fixed thus cannot meet various requirements applications. In this paper, we first introduce truncated Huber penalty function which shows strong flexibility under parameter settings. A generalized framework then proposed with introduced function. When combined its...

Video saliency detection aims to pop out the most salient regions in every frame of a video. Up now, many efforts have been made from various aspects for video detection. Unfortunately, existing models are very likely fail challenging videos with complicated motions and complex scenes. Therefore, this paper, we propose novel framework improve results generated by models. The proposed consists three key steps including localized estimation, spatiotemporal refinement, update. Specifically,...

Edge-preserving smoothing is a fundamental procedure for many computer vision and graphic applications. This can be achieved with either local methods or global methods. In most cases, yield superior performance over the ones. However, usually run much faster than this paper, we propose new method that embeds bilateral filter (BLF) in least squares (LS) model efficient edge-preserving smoothing. The proposed show comparable state-of-the-art method. Meanwhile, since take advantages of...

Salient object detection (SOD), which aims to identify and locate the most salient pixels or regions in images, has been attracting more interest due its various real-world applications. However, this vision task is quite challenging, especially under complex image scenes. Inspired by intrinsic reflection of natural paper we propose a novel feature learning framework for large-scale detection. Specifically, design symmetrical fully convolutional network (SFCN) effectively learn complementary...

Since multimodal learning is able to take advantage of the complementarity signals, performance emotion recognition usually surpasses that based on a single modality. In this paper, we introduce deep generalized canonical correlation analysis with an attention mechanism (DGCCA-AM) recognition. This model extends conventional (CCA) from two modalities arbitrarily numerous and implements adaptive fusion mechanism. By adjusting weights matrices maximize different modalities, DGCCA-AM extracts...

Image restoration includes various kinds of tasks, such as image denoising, deraining and low-light enhancement, etc. Due to the domain shift problem current supervised methods, researchers tend adopt unsupervised methods. However, fake color or blur image, insufficient missing semantic information are three common problems when utilizing these In this paper, we propose a new hybrid loss named Quality-Task-Perception (QTP) deal with simultaneously. Specifically, components: quality, task...

Salient Object Detection (SOD) aims to identify and segment the most conspicuous objects in an image or video. As important pre-processing step, it has many potential applications multimedia vision tasks. With advance of imaging devices, SOD with high-resolution images is great demand, recently. However, traditional methods are largely limited low-resolution images, making them difficult adapt development High-Resolution (HRSOD). Although some HRSOD emerge, there no large enough datasets for...

Introduction The poultry industry constantly seeks strategies to enhance broiler growth performance and overall health. Organic acidifiers, including L-lactic acid, L-malic acetic have gained attention as potential feed additives improve animal production by modulating gut health, enhancing nutrient absorption, supporting immune function. Despite their promising effects in other species, the impact of this novel compound organic acidifier on performance, metabolism, response has not been...

Solving the global method of Weighted Least Squares (WLS) model in image filtering is both time- and memory-consuming. In this paper, we present an alternative approximation a memory- efficient manner which denoted as Semi-Global Weighed (SG-WLS). Instead solving large linear system, propose to iteratively solve sequence subsystems are one-dimensional WLS models. Although each subsystem one-dimensional, it can take two-dimensional neighborhood information into account due proposed special...

Street Scene Parsing (SSP) is a fundamental and important step for autonomous driving traffic scene understanding. Recently, Fully Convolutional Network (FCN) based methods have delivered expressive performances with the help of large-scale dense-labeling datasets. However, in urban environments, not all labels contribute equally making control decision. Certain such as pedestrian, car, bicyclist, road lane or sidewalk would be more comparison vegetation, sky building. Based on this fact,...

In recent years, Salient Object Detection (SOD) has shown great success with the achievements of large-scale benchmarks and deep learning techniques. However, existing SOD methods mainly focus on natural images low-resolutions, e.g., $400\times 400$ or less. This drawback hinders them for advanced practical applications, which need high-resolution, detail-aware results. Besides, lacking boundary detail semantic context salient objects is also a key concern accurate SOD. To address these...

Salient object detection, which aims to identify and locate the most salient pixels or regions in images, has been attracting more interest due its various real-world applications. However, this vision task is quite challenging, especially under complex image scenes. Inspired by intrinsic reflection of natural paper we propose a novel feature learning framework for large-scale detection. Specifically, design symmetrical fully convolutional network (SFCN) learn complementary saliency features...