Filters








38 Hits in 3.9 sec

MAT: Mask-Aware Transformer for Large Hole Image Inpainting [article]

Wenbo Li, Zhe Lin, Kun Zhou, Lu Qi, Yi Wang, Jiaya Jia
2022 arXiv   pre-print
In this paper, we present a novel transformer-based model for large hole inpainting, which unifies the merits of transformers and convolutions to efficiently process high-resolution images.  ...  Specifically, we customize an inpainting-oriented transformer block, where the attention module aggregates non-local information only from partial valid tokens, indicated by a dynamic mask.  ...  Conclusion We have presented a mask-aware transformer (MAT) for pluralistic large hole image inpainting.  ... 
arXiv:2203.15270v2 fatcat:7zeh7kmo3farpihku6o62oryem

AutoRemover: Automatic Object Removal for Autonomous Driving Videos

Rong Zhang, Wei Li, Peng Wang, Chenye Guan, Jin Fang, Yuhang Song, Jinhui Yu, Baoquan Chen, Weiwei Xu, Ruigang Yang
2020 PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE AND THE TWENTY-EIGHTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE  
Motivated by the need for photo-realistic simulation in autonomous driving, in this paper we present a video inpainting algorithm AutoRemover, designed specifically for generating street-view videos without  ...  To deal with large ego-motion, we take advantage of the multi-source data, in particular the 3D data, in autonomous driving.  ...  Acknowledgements Weiwei Xu is partially supported by NSFC (No. 61732016) and the fundamental research fund for the central universities. Jinhui Yu is partially supported by NSFC (No. 61772463).  ... 
doi:10.1609/aaai.v34i07.6982 fatcat:kgrz7qzw6zhmrcsdoearrzzwna

AutoRemover: Automatic Object Removal for Autonomous Driving Videos [article]

Rong Zhang, Wei Li, Peng Wang, Chenye Guan, Jin Fang, Yuhang Song, Jinhui Yu, Baoquan Chen, Weiwei Xu, Ruigang Yang
2019 arXiv   pre-print
Motivated by the need for photo-realistic simulation in autonomous driving, in this paper we present a video inpainting algorithm AutoRemover, designed specifically for generating street-view videos without  ...  To deal with large ego-motion, we take advantage of the multi-source data, in particular the 3D data, in autonomous driving.  ...  Acknowledgements Weiwei Xu is partially supported by NSFC (No. 61732016) and the fundamental research fund for the central universities. Jinhui Yu is partially supported by NSFC (No. 61772463).  ... 
arXiv:1911.12588v1 fatcat:223s5gmqjbczzcccbb6pt64neq

Depth-Guided Disocclusion Inpainting of Synthesized RGB-D Images

Pierre Buyssens, Olivier Le Meur, Maxime Daisy, David Tschumperle, Olivier Lezoray
2017 IEEE Transactions on Image Processing  
We propose to tackle the disocclusion inpainting of RGB-D images when synthesizing new views of a scene by changing its viewpoint. Such a process creates holes both in depth and color images.  ...  Relevant comparisons to state-of-the-art inpainting methods for the disocclusion inpainting of both depth and color images are provided and illustrate the effectiveness of our proposed algorithms.  ...  In section IV, we detail our depth-aware patch-based approach for color image disocclusion inpainting.  ... 
doi:10.1109/tip.2016.2619263 pmid:27775518 fatcat:nqlgadxfdne5bcmvgd3pv2gt5q

Unselfie: Translating Selfies to Neutral-pose Portraits in the Wild [article]

Liqian Ma, Zhe Lin, Connelly Barnes, Alexei A. Efros, Jingwan Lu
2020 arXiv   pre-print
To achieve this, we first collect an unpaired dataset, and introduce a way to synthesize paired training data for self-supervised learning.  ...  To address this issue, we introduce unselfie, a novel photographic transformation that automatically translates a selfie into a neutral-pose portrait.  ...  We thank He Zhang for helping mask estimation.  ... 
arXiv:2007.15068v1 fatcat:nijdne54znfj3jvccuylhzkeua

Learning to Remove Soft Shadows

Maciej Gryka, Michael Terry, Gabriel J. Brostow
2015 ACM Transactions on Graphics  
We tested the approach on a large set of soft shadow images, and performed a user study that compared our method to the state of the art and to real lit scenes.  ...  We posit that soft shadows can be segmented, and therefore edited, by learning a mapping function for image patches that generates shadow mattes.  ...  Gabriel Brostow is grateful for the UK EP-  ... 
doi:10.1145/2732407 fatcat:q76up74q3bdiff7v7nw5g73ava

NeuralReshaper: Single-image Human-body Retouching with Deep Neural Networks [article]

Beijia Chen, Hongbo Fu, Xiang Chen, Kun Zhou, Youyi Zheng
2022 arXiv   pre-print
In this paper, we present NeuralReshaper, a novel method for semantic reshaping of human bodies in single images using deep generative networks.  ...  Previous methods rely on image warping to transfer 3D reshaping effects to the entire image domain and thus often cause distortions in both foreground and background.  ...  Early image inpainting methods employ diffusion techniques [5, 9, 38] or patch match [6, 7, 15, 16, 36, 63, 68] to fill small holes based on low-level image features, having limited ability for hallucinating  ... 
arXiv:2203.10496v2 fatcat:zn65plymgvahnm4exq7stwsln4

Deep Fusion Network for Image Completion [article]

Xin Hong and Pengfei Xiong and Renhe Ji and Haoqiang Fan
2019 arXiv   pre-print
Deep image completion usually fails to harmonically blend the restored image into existing content, especially in the boundary area.  ...  Firstly, a fusion block is introduced to generate a flexible alpha composition map for combining known and unknown regions.  ...  For PConv, we think this is the limit of partial convolution, which can't transmit information into a very large hole.  ... 
arXiv:1904.08060v1 fatcat:m57hk6jc6je2xday3gcaedarsi

Deep Depth Completion of a Single RGB-D Image [article]

Yinda Zhang, Thomas Funkhouser
2018 arXiv   pre-print
., inpainting depths directly) as the result of extensive experiments with a new depth completion benchmark dataset, where holes are filled in training data through the rendering of surface reconstructions  ...  The goal of our work is to complete the depth channel of an RGB-D image. Commodity-grade depth cameras often fail to sense depth for shiny, bright, transparent, and distant surfaces.  ...  " depth images (e.g., where ground-truth depth is provided for holes).  ... 
arXiv:1803.09326v2 fatcat:jezuunhe6fearnnfx3544s3bnu

Sky Optimization: Semantically aware image processing of skies in low-light photography [article]

Orly Liba, Longqi Cai, Yun-Ta Tsai, Elad Eban, Yair Movshovitz-Attias, Yael Pritch, Huizhong Chen, Jonathan T. Barron
2020 arXiv   pre-print
We introduce a method for creating an accurate sky-mask dataset that is based on partially annotated images that are inpainted and refined by our modified weighted guided filter.  ...  Our modified weighted guided filter is used for edge-aware upsampling to resize the alpha-mask to a higher resolution.  ...  Because of the inherent difficulty of manually annotating a pixel-level alpha for a high-detailed sky mask, we propose an algorithmic approach for transforming approximate binary segmentation masks into  ... 
arXiv:2006.10172v1 fatcat:bv7ejd2q75aszk2yq3wnbfduwu

SAC-GAN: Structure-Aware Image-to-Image Composition for Self-Driving [article]

Hang Zhou, Ali Mahdavi-Amiri, Rui Ma, Hao Zhang
2022 arXiv   pre-print
We evaluate our network, coined SAC-GAN for structure-aware composition, on prominent self-driving datasets in terms of quality, composability, and generalizability of the composite images.  ...  We present a compositional approach to image augmentation for self-driving applications.  ...  Note that the background scene is obtained by taking out the foreground object and performing inpainting to fill the holes.  ... 
arXiv:2112.06596v3 fatcat:6ftcz32xxnalhpneektwpxuxgi

MichiGAN: Multi-Input-Conditioned Hair Image Generation for Portrait Editing [article]

Zhentao Tan and Menglei Chai and Dongdong Chen and Jing Liao and Qi Chu and Lu Yuan and Sergey Tulyakov and Nenghai Yu
2020 arXiv   pre-print
In this paper, we present MichiGAN (Multi-Input-Conditioned Hair Image GAN), a novel conditional image generation method for interactive portrait hair manipulation.  ...  For each of them, we design a corresponding condition module to represent, process, and convert user inputs, and modulate the image generation pipeline in ways that respect the natures of different visual  ...  acknowledgements We would like to thank the reviewers for their constructive feedback, Mingming He and Jian Ren for helpful discussions, Nvidia Research for making the Flickr-Faces-HQ (FFHQ) dataset, and  ... 
arXiv:2010.16417v1 fatcat:kwpceg2hfbfavbfydwsi6locwq

MPB: A modified Poisson blending technique

Mahmoud Afifi, Khaled F. Hussain
2015 Computational Visual Media  
This makes the proposed technique suitable for use in video compositing as it avoids the flickering caused by bleeding artifacts.  ...  Image cloning has many useful applications, such as removing unwanted objects, fixing damaged parts of images, and panorama stitching.  ...  Afifi for their cooperation for allowing use of their footage in the experimental results.  ... 
doi:10.1007/s41095-015-0027-z fatcat:d3iexs5bqffo7bh76bdbgos6mm

Transformation-Grounded Image Generation Network for Novel 3D View Synthesis

Eunbyung Park, Jimei Yang, Ersin Yumer, Duygu Ceylan, Alexander C. Berg
2017 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)  
We present a transformation-grounded image generation network for novel 3D view synthesis from a single image.  ...  In addition to the new network structure, training with a combination of adversarial and perceptual loss results in a reduction in common artifacts of novel view synthesis such as distortions and holes  ...  We would like to thank Weilin Sun, Guilin Liu, True Price, and Dinghuang Ji for helpful discussions. We thank NVIDIA for providing GPUs and acknowledge support from NSF 1452851, 1526367.  ... 
doi:10.1109/cvpr.2017.82 dblp:conf/cvpr/ParkYYCB17 fatcat:ltj2uqgrvje65lpo7h2fozbdze

Transformation-Grounded Image Generation Network for Novel 3D View Synthesis [article]

Eunbyung Park, Jimei Yang, Ersin Yumer, Duygu Ceylan, Alexander C. Berg
2017 arXiv   pre-print
We present a transformation-grounded image generation network for novel 3D view synthesis from a single image.  ...  In addition to the new network structure, training with a combination of adversarial and perceptual loss results in a reduction in common artifacts of novel view synthesis such as distortions and holes  ...  We would like to thank Weilin Sun, Guilin Liu, True Price, and Dinghuang Ji for helpful discussions. We thank NVIDIA for providing GPUs and acknowledge support from NSF 1452851, 1526367.  ... 
arXiv:1703.02921v1 fatcat:acvd3262zrg7ra7rwajo3tg65q
« Previous Showing results 1 — 15 out of 38 results