Filters








1,447 Hits in 4.5 sec

Increasing Video Perceptual Quality with GANs and Semantic Coding

Leonardo Galteri, Marco Bertini, Lorenzo Seidenari, Tiberio Uricchio, Alberto Del Bimbo
2020 Proceedings of the 28th ACM International Conference on Multimedia  
KEYWORDS semantic video compression, GANs, video quality enhancement ACM Reference Format:  ...  Our study shows that the combination of semantic coding and learning based video restoration can provide superior results.  ...  ACKNOWLEDGMENTS We gratefully acknowledge the support of NVIDIA Corporation with the donation of the Titan Xp GPU used for this research.  ... 
doi:10.1145/3394171.3413508 dblp:conf/mm/GalteriBSUB20 fatcat:vmtyaf7aojbhfbaic7777zmcce

Feature-Style Encoder for Style-Based GAN Inversion [article]

Xu Yao, Alasdair Newson, Yann Gousseau, Pierre Hellier
2022 arXiv   pre-print
Our model achieves accurate inversion of real images from the latent space of a pre-trained style-based GAN model, obtaining better perceptual quality and lower reconstruction error than existing methods  ...  Additionally, we demonstrate that the proposed encoder is especially well-suited for inversion and editing on videos.  ...  However, compared with the synthetic images, the perceptual quality of the images resulting from the inversion is worse.  ... 
arXiv:2202.02183v1 fatcat:7xtfszrk6vhgxavmz7zu2m2pfm

Perceptual Indistinguishability-Net (PI-Net): Facial Image Obfuscation with Manipulable Semantics [article]

Jia-Wei Chen, Li-Ju Chen, Chia-Mu Yu, Chun-Shien Lu
2021 arXiv   pre-print
In this study, with the consideration of the perceptual similarity, we propose perceptual indistinguishability (PI) as a formal privacy notion particularly for images.  ...  With the growing use of camera devices, the industry has many image datasets that provide more opportunities for collaboration between the machine learning community and industry.  ...  Unfortunately, as video adjacency is defined in pixel domain, the video quality will be largely destroyed.  ... 
arXiv:2104.01753v2 fatcat:vdbhpaj7sje2plfdftqi7rggsu

Video Coding Using Learned Latent GAN Compression [article]

Mustafa Shukor, Bharath Bhushan Damodaran, Xu Yao, Pierre Hellier
2022 arXiv   pre-print
We leverage the generative capacity of GANs such as StyleGAN to represent and compress a video, including intra and inter compression.  ...  Finally, an entropy model for video inter coding with residual is also learned in the previously constructed latent representation.  ...  We propose to learn a model for intra and inter video coding. • We propose a new perceptual distortion loss that is more efficient to compute and leverages the multiscale and semantic representation in  ... 
arXiv:2207.04324v2 fatcat:67afrd5xm5ewbdi2wyefxr6ajy

Adversarial Video Compression Guided by Soft Edge Detection [article]

Sungsoo Kim and Jin Soo Park and Christos G. Bampis and Jaeseong Lee and Mia K. Markey and Alexandros G. Dimakis and Alan C. Bovik
2018 arXiv   pre-print
Experiments on a diverse set of 131 videos demonstrate that our proposed GAN-based compression engine achieves much higher quality reconstructions at very low bitrates than prevailing standard codecs such  ...  We propose a video compression framework using conditional Generative Adversarial Networks (GANs).  ...  traits of latent codes [10] , to build a semantic relationship between a latent code and an output code.  ... 
arXiv:1811.10673v1 fatcat:z4bkbnxg3bcjjmpzvvspfmpzuq

Generative Compression

Shibani Santurkar, David Budden, Nir Shavit
2018 2018 Picture Coding Symposium (PCS)  
reconstructions at much deeper compression levels for both image and video data.  ...  Traditional image and video compression algorithms rely on hand-crafted encoder/decoder pairs (codecs) that lack adaptability and are agnostic to the data being compressed.  ...  Acknowledgments Support is gratefully acknowledged from the National Science Foundation (NSF) under grants IIS-1447786 and CCF-1563880, and the Intelligence Advanced Research Projects Activity (IARPA)  ... 
doi:10.1109/pcs.2018.8456298 dblp:conf/pcs/SanturkarBS18 fatcat:b564wxlorrat3owipbx4422tda

Generative Compression [article]

Shibani Santurkar, David Budden, Nir Shavit
2017 arXiv   pre-print
reconstructions at much deeper compression levels for both image and video data.  ...  Traditional image and video compression algorithms rely on hand-crafted encoder/decoder pairs (codecs) that lack adaptability and are agnostic to the data being compressed.  ...  Acknowledgments Support is gratefully acknowledged from the National Science Foundation (NSF) under grants IIS-1447786 and CCF-1563880, and the Intelligence Advanced Research Projects Activity (IARPA)  ... 
arXiv:1703.01467v2 fatcat:mpkm5tvw7jdh5iyl3lblxps3ai

Adversarial Distortion for Learned Video Compression

Vijay Veerabadran, Reza Pourreza, Amirhossein Habibian, Taco Cohen
2020 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)  
We find this adversarial objective to correlate better with human perceptual quality judgement relative to traditional quality metrics such as MS-SSIM and PSNR.  ...  Our experiments using a state-of-the-art learned video compression system demonstrate a reduction of perceptual artifacts and reconstruction of detail lost especially under extremely high compression.  ...  Recent work [6] mathematically proves that the distortion and perceptual quality are at odds with each other and minimizing the mean distortion leads to a decrease in perceptual quality.  ... 
doi:10.1109/cvprw50498.2020.00092 dblp:conf/cvpr/VeerabadranPHC20 fatcat:crkibnytbre5zkg5vxwitfkyzi

Adversarial Distortion for Learned Video Compression [article]

Vijay Veerabadran, Reza Pourreza, Amirhossein Habibian, Taco Cohen
2021 arXiv   pre-print
We find this adversarial objective to correlate better with human perceptual quality judgement relative to traditional quality metrics such as MS-SSIM and PSNR.  ...  Our experiments using a state-of-the-art learned video compression system demonstrate a reduction of perceptual artifacts and reconstruction of detail lost especially under extremely high compression.  ...  Recent work [6] mathematically proves that the distortion and perceptual quality are at odds with each other and minimizing the mean distortion leads to a decrease in perceptual quality.  ... 
arXiv:2004.09508v3 fatcat:puoumoefdvdg7k5l3mly2rkdna

Image2StyleGAN++: How to Edit the Embedded Images? [article]

Rameen Abdal, Yipeng Qin, Peter Wonka
2020 arXiv   pre-print
Third, we combine embedding with activation tensor manipulation to perform high-quality local edits along with global semantic edits on images.  ...  Our noise optimization can restore high-frequency features in images and thus significantly improves the quality of reconstructed images, e.g. a big increase of PSNR from 20 dB to 45 dB.  ...  Introduction Recent GANs [18, 6] demonstrated that synthetic images can be generated with very high quality.  ... 
arXiv:1911.11544v2 fatcat:5ombxvb6nrdkno5ccj374zl3cq

Multi-level Wavelet-based Generative Adversarial Network for Perceptual Quality Enhancement of Compressed Video [article]

Jianyi Wang, Xin Deng, Mai Xu, Congyong Chen, Yuhang Song
2020 arXiv   pre-print
In this paper, we focus on enhancing the perceptual quality of compressed video.  ...  Existing methods mainly focus on enhancing the objective quality of compressed video while ignoring its perceptual quality.  ...  As the first attempt to enhance the perceptual quality of compressed video, our method achieves better perceptual quality with lower LPIPS and PI. Zoom in for best view. [32] .  ... 
arXiv:2008.00499v1 fatcat:owuqcqqakvfpnefkrne7cmbjye

Extreme Learned Image Compression with GANs

Eirikur Agustsson, Michael Tschannen, Fabian Mentzer, Radu Timofte, Luc Van Gool
2018 Computer Vision and Pattern Recognition  
Ours (0.036bpp) BPG (0.039bpp) Figure 1 : Images produced by our global generative compression network trained with an adversarial loss, along with the corresponding results for BPG.  ...  Therefore, to quantitatively evaluate the perceptual quality of our GC networks in comparison with BPG and AEDC we conduct a user study using the Amazon Mechanical Turk (AMT) platform.  ...  While there is some color shift noticeable (which could be accounted for by reducing the domain mismatch and/or increasing the weight of the perceptual loss), we see that our method can realistically synthesize  ... 
dblp:conf/cvpr/AgustssonTMTG18 fatcat:zp2u3sqvmbae7n2abbnpioqbhi

Image2StyleGAN++: How to Edit the Embedded Images?

Rameen Abdal, Yipeng Qin, Peter Wonka
2020 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)  
Figure 1: (a) and (b): input images; (c): the "two-face" generated by naively copying the left half from (a) and the right half from (b); (d): the "two-face" generated by our Image2StyleGAN++ framework  ...  Introduction Recent GANs [19, 6] demonstrated that synthetic images can be generated with very high quality.  ...  We investigate the combination of embedding and activation tensor manipulation to perform high quality local edits along with global semantic edits on images. 4.  ... 
doi:10.1109/cvpr42600.2020.00832 dblp:conf/cvpr/AbdalQW20 fatcat:56k57eurjfha3nj3hvn6rhylda

Image2StyleGAN: How to Embed Images Into the StyleGAN Latent Space? [article]

Rameen Abdal, Yipeng Qin, Peter Wonka
2019 arXiv   pre-print
We propose a set of experiments to test what class of images can be embedded, how they are embedded, what latent space is suitable for embedding, and if the embedding is semantically meaningful.  ...  This embedding enables semantic image editing operations that can be applied to existing photographs.  ...  In the few past years, the quality of images synthesized by GANs has increased rapidly.  ... 
arXiv:1904.03189v2 fatcat:4s52z5vxwffhjhtxbjxjmpzxke

HeadNeRF: A Real-time NeRF-based Parametric Head Model [article]

Yang Hong, Bo Peng, Haiyao Xiao, Ligang Liu, Juyong Zhang
2022 arXiv   pre-print
It can render high fidelity head images in real-time on modern GPUs, and supports directly controlling the generated images' rendering pose and various semantic attributes.  ...  The well designed loss terms also improve the rendering accuracy, and the fine-level details of the human head, such as the gaps between teeth, wrinkles, and beards, can be represented and synthesized  ...  codes from the reference video.  ... 
arXiv:2112.05637v3 fatcat:ikjhucsrpva7rnysuzkz3zg65i
« Previous Showing results 1 — 15 out of 1,447 results