Image inpainting techniques have shown significant improvements by using deep neural networks recently. However, most of them may either fail to reconstruct reasonable structures or restore fine-grained textures. In order solve this problem, in paper, we propose a two-stage model which splits the task into two parts: structure reconstruction and texture generation. first stage, edge-preserved smooth images are employed train reconstructor completes missing inputs. second based on...

Low-light images are not conducive to human observation and computer vision algorithms due their low visibility. To solve this problem, many image enhancement techniques have been proposed. However, existing inevitably introduce color lightness distortion when increasing lower the distortion, we propose a novel method using response characteristics of cameras. First, investigate relationship between two with different exposures obtain an accurate camera model. Then borrow illumination...

Low-light image enhancement algorithms can improve the visual quality of low-light images and support extraction valuable information for some computer vision techniques. However, existing techniques inevitably introduce color lightness distortions when enhancing images. To lower distortions, we propose a novel framework using response characteristics cameras. First, discuss how to determine reasonable camera model its parameters. Then, use illumination estimation estimate exposure ratio...

Pose-guided person image generation is to transform a source target pose. This task requires spatial manipulations of data. However, Convolutional Neural Networks are limited by the lack ability spatially inputs. In this paper, we propose differentiable global-flow local-attention framework reassemble inputs at feature level. Specifically, our model first calculates global correlations between sources and targets predict flow fields. Then, flowed local patch pairs extracted from maps...

Generating portrait images by controlling the motions of existing faces is an important task great consequence to social media industries. For easy use and intuitive control, semantically meaningful fully disentangled parameters should be used as modifications. However, many techniques do not provide such fine-grained controls or indirect editing methods i.e. mimic other individuals. In this paper, a Portrait Image Neural Renderer (PIRenderer) proposed control face with three-dimensional...

We deal with the controllable person image synthesis task which aims to re-render a human from reference explicit control over body pose and appearance. Observing that images are highly structured, we propose generate desired by extracting distributing semantic entities of images. To achieve this goal, neural texture extraction distribution operation based on double attention is described. This first extracts textures feature maps. Then, it distributes extracted according spatial...

Pose-guided person image generation and animation aim to transform a source target poses. These tasks require spatial manipulation of data. However, Convolutional Neural Networks are limited by the lack ability spatially inputs. In this article, we propose differentiable global-flow local-attention framework reassemble inputs at feature level. This first estimates global flow fields between sources targets. Then, corresponding local patches sampled with content-aware attention coefficients....

Image aesthetic assessment involves both fine-grained details and the holistic layout of images. However, most current approaches learn local information separately, which has a potential loss contextual information. Additionally, learning-based methods mainly cast image as binary classification or regression problem, cannot sufficiently delineate diversity human experience. To address these limitations, we attempt to render model varieties Specifically, explore context-aware attention...

Pose-guided person image generation is to transform a source target pose. This task requires spatial manipulations of data. However, Convolutional Neural Networks are limited by the lack ability spatially inputs. In this paper, we propose differentiable global-flow local-attention framework reassemble inputs at feature level. Specifically, our model first calculates global correlations between sources and targets predict flow fields. Then, flowed local patch pairs extracted from maps...

Point cloud compression plays a crucial role in reducing the huge cost of data storage and transmission. However, distortions can be introduced into decompressed point clouds due to quantization. In this paper, we propose novel learning-based post-processing method enhance clouds. Specifically, voxelized is first divided small cubes. Then, 3D convolutional network proposed predict occupancy probability for each location cube. We leverage both local global contexts by generating multi-scale...

Image inpainting techniques have shown significant improvements by using deep neural networks recently. However, most of them may either fail to reconstruct reasonable structures or restore fine-grained textures. In order solve this problem, in paper, we propose a two-stage model which splits the task into two parts: structure reconstruction and texture generation. first stage, edge-preserved smooth images are employed train reconstructor completes missing inputs. second based on...

Pose-guided person image synthesis aims to synthesize images by transforming reference into target poses. In this paper, we observe that the commonly used spatial transformation blocks have complementary advantages. We propose a novel model combining attention operation with flow-based operation. Our not only takes advantage of generate accurate structures but also uses sample realistic source textures. Both objective and subjective experiments demonstrate superiority our model. Meanwhile,...

Point cloud completion is the task of estimating complete point from partial observation. Most existing methods tend to recover global shapes 3D objects and usually lack local details. These rely merely on distance metrics between sets as loss functions, which have insufficient capability supervising fine structures. In this work, we propose a coarse-to-fine approach with two stages: 1) Flow-based Completion Network, principled probabilistic model that built continuous normalizing flow...

The things of missed assembling, reversed assembling and mixed the pistons happen from time to time. assemble quality has severe influence on performance vehicle engine. With aid machine vision recognition technology, these mistakes could be successfully avoided; it makes piston stable labor intensity lower.

Pose-guided person image generation aims to transfer reference images target poses while preserving the source appearance. Recent approaches achieve considerable improvement by using spatial transformation modules such as attention operation. However, commonly used vanilla tends generate a dense correlation matrix which means that value of position is weighted sum many positions, resulting in blurry In this paper, we propose novel model named Flow-guided Attention Deformation (FAD) perform...

We deal with the controllable person image synthesis task which aims to re-render a human from reference explicit control over body pose and appearance. Observing that images are highly structured, we propose generate desired by extracting distributing semantic entities of images. To achieve this goal, neural texture extraction distribution operation based on double attention is described. This first extracts textures feature maps. Then, it distributes extracted according spatial...

Point cloud compression plays a crucial role in reducing the huge cost of data storage and transmission. However, distortions can be introduced into decompressed point clouds due to quantization. In this paper, we propose novel learning-based post-processing method enhance clouds. Specifically, voxelized is first divided small cubes. Then, 3D convolutional network proposed predict occupancy probability for each location cube. We leverage both local global contexts by generating multi-scale...

Video-Text Retrieval aims to perform accurate retrieval process that adopts texts retrieve the corresponding videos, and vice versa. Typically, mainstream methods solve this problem by learning a common joint embedding space, then measure similarities between videos texts. However, these lack ability represent detailed semantic information. Therefore, we first utilize three pre-trained models construct video embeddings of different levels, propose Context-aware Hierarchical Transformer (CHT)...

Pose-guided person image synthesis aims to synthesize images by transforming reference into target poses. In this paper, we observe that the commonly used spatial transformation blocks have complementary advantages. We propose a novel model combining attention operation with flow-based operation. Our not only takes advantage of generate accurate structures but also uses sample realistic source textures. Both objective and subjective experiments demonstrate superiority our model. Meanwhile,...

Generating portrait images by controlling the motions of existing faces is an important task great consequence to social media industries. For easy use and intuitive control, semantically meaningful fully disentangled parameters should be used as modifications. However, many techniques do not provide such fine-grained controls or indirect editing methods i.e. mimic other individuals. In this paper, a Portrait Image Neural Renderer (PIRenderer) proposed control face with three-dimensional...