Filters








2,280 Hits in 7.4 sec

Variable Rate Deep Image Compression with Modulated Autoencoder [article]

Fei Yang, Luis Herranz, Joost van de Weijer, José A. Iglesias Guitián, Antonio López, Mikhail Mozerov
2019 arXiv   pre-print
Variable rate is a requirement for flexible and adaptable image and video compression. However, deep image compression methods are optimized for a single fixed rate-distortion tradeoff.  ...  Addressing these limitations, we formulate the problem of variable rate-distortion optimization for deep image compression, and propose modulated autoencoders (MAEs), where the representations of a shared  ...  MULTI-RATE DEEP IMAGE COMPRESSION WITH MODULATED AUTOENCODERS A.  ... 
arXiv:1912.05526v1 fatcat:4bxstnwrnnerpp4efatssyzpoa

Lossy Medical Image Compression using Residual Learning-based Dual Autoencoder Model [article]

Dipti Mishra, Satish Kumar Singh, Rajat Kumar Singh
2021 arXiv   pre-print
In this work, we propose a two-stage autoencoder based compressor-decompressor framework for compressing malaria RBC cell image patches.  ...  The proposed residual-based dual autoencoder network is trained to extract the unique features which are then used to reconstruct the original image through the decompressor module.  ...  ROI implies that the different parts of the image can be encoded with variable bit rates providing variable image qualities.  ... 
arXiv:2108.10579v1 fatcat:2no5y3hvnvf4tbu3zyawmua4vq

G-VAE: A Continuously Variable Rate Deep Image Compression Framework [article]

Ze Cui, Jing Wang, Bo Bai, Tiansheng Guo, Yihui Feng
2020 arXiv   pre-print
Rate adaption of deep image compression in a single model will become one of the decisive factors competing with the classical image compression codecs.  ...  In this paper, we propose a novel image compression framework G-VAE (Gained Variational Autoencoder), which could achieve continuously variable rate in a single model.  ...  [29] proposed a variable rate image compression framework with a conditional autoencoder, which incorporates fully connection networks into the convolution unit and adjusts compression performance with  ... 
arXiv:2003.02012v2 fatcat:nhoxz55pkjh4dlxuh3greand2a

CAESR: Conditional Autoencoder and Super-Resolution for Learned Spatial Scalability [article]

Charles Bonnineau, Wassim Hamidouche, Jean-François Travers, Naty Sidaty, Jean-Yves Aubié, Olivier Deforges
2022 arXiv   pre-print
Our framework considers a low-resolution signal encoded with VVC intra-mode as a base-layer (BL), and a deep conditional autoencoder with hyperprior (AE-HP) as an enhancement-layer (EL) model.  ...  On the decoder side, a super-resolution (SR) module is used to recover high-resolution details and invert the conditional coding process.  ...  On the other hand, end-to-end learning models for image and video compression were proposed using deep autoencoders (AEs) [8] - [12] .  ... 
arXiv:2202.00416v1 fatcat:5ilwqyfulrbfvaavf3e7e2y3ja

Scalable Recollections for Continual Lifelong Learning [article]

Matthew Riemer, Tim Klinger, Djallel Bouneffouf, Michele Franceschini
2018 arXiv   pre-print
Given the recent success of Deep Learning applied to a variety of single tasks, it is natural to consider more human-realistic settings.  ...  Deep non-linear autoencoders are a natural choice for compression problems.  ...  The categorical latent variable autoencoders explored had the following representation sizes: 168 2d variables for 10x compression, 62 2d variables for 50x compression, and 38 2d variables for 100x compression  ... 
arXiv:1711.06761v4 fatcat:4sblnpp7n5hrlo2wioqzx5au44

Variable Rate Image Compression Method with Dead-zone Quantizer [article]

Jing Zhou, Akira Nakagawa, Keizo Kato, Sihan Wen, Kimihiko Kazui, Zhiming Tan
2020 arXiv   pre-print
For conventional codec, signal is decorrelated with orthonmal transformation, and uniform quantizer is introduced. We propose a variable rate image compression method with dead-zone quantizer.  ...  Deep learning based image compression methods have achieved superior performance compared with transform based conventional codec.  ...  Variable rate image compression Training framework: RaDOGAGA As for RaDOGAGA, which is a rate-distortion optimization guided autoencoder, it proves that deep autoencoder can achieve orthonormal transform  ... 
arXiv:2004.05855v2 fatcat:r2gtorairbeltpz6lyizlefxhm

Slimmable Compressive Autoencoders for Practical Neural Image Compression [article]

Fei Yang, Luis Herranz, Yongmei Cheng, Mikhail G. Mozerov
2022 arXiv   pre-print
Neural image compression leverages deep neural networks to outperform traditional image codecs in rate-distortion performance.  ...  Focusing on practical image compression, we propose slimmable compressive autoencoders (SlimCAEs), where rate (R) and distortion (D) are jointly optimized for different capacities.  ...  Figure 1 : 1 Figure 1: Variable rate and complexity adaptive image compression with a slimmable compressive autoencoder.  ... 
arXiv:2103.15726v2 fatcat:5wrayrmvnjedxink26lpnuuph4

DeepSIC: Deep Semantic Image Compression [article]

Sihui Luo, Yezhou Yang, Mingli Song
2018 arXiv   pre-print
In this paper, we propose a concept called Deep Semantic Image Compression (DeepSIC) and put forward two novel architectures that aim to reconstruct the compressed image and generate corresponding semantic  ...  In both architectures, the feature maps are shared by the compression and the semantic analytics modules.  ...  DEEP SEMANTIC IMAGE COMPRESSION In the proposed deep semantic image compression framework (DeepSIC), the compression scheme is similar to autoencoder.  ... 
arXiv:1801.09468v1 fatcat:rbt6hpixlvchficu6vrbucljhi

Deep Joint Source-channel Coding for Wireless Image Transmission

Eirina Bourtsoulatze, David Burth Kurka, Deniz Gunduz
2019 ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)  
Our results show that the proposed deep JSCC scheme outperforms digital transmission concatenating JPEG or JPEG2000 compression with a capacity achieving channel code at low signal-tonoise ratio (SNR)  ...  Index Terms-Joint source-channel coding, deep neural networks, image communications.  ...  source compression with JPEG/JPEG2000 followed by channel coding and modulation, is inferior to the performance of the proposed deep JSCC.  ... 
doi:10.1109/icassp.2019.8683463 dblp:conf/icassp/BourtsoulatzeKG19 fatcat:ujmddculwvgivoi7grxqibsi5a

End-to-end Optimized Video Compression with MV-Residual Prediction [article]

XiangJi Wu, Ziwen Zhang, Jie Feng, Lei Zhou, Junmin Wu
2020 arXiv   pre-print
The prior probability of the latent representations is modeled by a hyperprior autoencoder and trained jointly with the MV-Residual network.  ...  Finally, novel rate allocation and post-processing strategies are used to produce the final compressed bits, considering the bits constraint of the challenge.  ...  In order to make full use of every bit space available, a rate control module [4] is utilized to fit our model to variable compression rates with a single set of weights.  ... 
arXiv:2005.12945v1 fatcat:xcibnnubn5dqnm42j2juot7iri

Learned Variable-Rate Multi-Frequency Image Compression using Modulated Generalized Octave Convolution [article]

Jianping Lin, Mohammad Akbari, Haisheng Fu, Qian Zhang, Shang Wang, Jie Liang, Dong Liu, Feng Liang, Guohe Zhang, Chengjie Tu
2020 arXiv   pre-print
In addition, we develop a variable-rate scheme using the Lagrangian parameter to modulate all the internal feature maps in the auto-encoder, which allows the scheme to achieve the large bitrate range of  ...  In this proposal, we design a learned multi-frequency image compression approach that uses generalized octave convolutions to factorize the latent representations into high-frequency (HF) and low-frequency  ...  Index Terms-learned image compression, octave convolution, variable-rate deep learning models, modulated scheme I. METHOD A.  ... 
arXiv:2009.13074v1 fatcat:z7w24y6y2bhybcjwfhtwhjvmva

Deep Joint Source-Channel Coding for Wireless Image Transmission [article]

Eirina Bourtsoulatze, David Burth Kurka, Deniz Gunduz
2019 arXiv   pre-print
Our results show that the proposed deep JSCC scheme outperforms digital transmission concatenating JPEG or JPEG2000 compression with a capacity achieving channel code at low signal-to-noise ratio (SNR)  ...  We propose a joint source and channel coding (JSCC) technique for wireless image transmission that does not rely on explicit codes for either compression or error correction; instead, it directly maps  ...  source compression with JPEG/JEPG2000 followed by channel coding and modulation, is inferior to the performance of the proposed deep JSCC.  ... 
arXiv:1809.01733v4 fatcat:pradboay7na27cihy4kh6holva

DRASIC: Distributed Recurrent Autoencoder for Scalable Image Compression [article]

Enmao Diao, Jie Ding, Vahid Tarokh
2019 arXiv   pre-print
The proposed architecture, which we refer to as Distributed Recurrent Autoencoder for Scalable Image Compression (DRASIC), is able to train distributed encoders and one joint decoder on correlated data  ...  We propose a new architecture for distributed image compression from a group of distributed data sources.  ...  Image compression with Deep Learning There exist a variety of classical codecs for lossy image compression.  ... 
arXiv:1903.09887v3 fatcat:pggenmvw65cvvinu5fo2eh4wmy

Scalable Recollections for Continual Lifelong Learning

Matthew Riemer, Tim Klinger, Djallel Bouneffouf, Michele Franceschini
2019 PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE AND THE TWENTY-EIGHTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE  
Given the recent success of Deep Learning applied to a variety of single tasks, it is natural to consider more human-realistic settings.  ...  We have shown that VAEs with categorical latent variables significantly outperform those with continuous latent variables (and even JPEG) for lossy compression.  ...  In Figure 3 we empirically demonstrate that autoencoders with categorical latent variables can achieve significantly more storage compression of input observations at the same average distortion as autoencoders  ... 
doi:10.1609/aaai.v33i01.33011352 fatcat:wnxdvhxyojf4xepljtua3565j4

Churn Prediction using Attention Based Autoencoder Network

Swetha Allam, Velagapudi Ramakrishna Siddhartha Engineering College, India
2019 International Journal of Advanced Trends in Computer Science and Engineering  
To overcome these problems, in this paper, we are proposing a feature extraction model based on Autoencoder with attention mechanism.  ...  Effective churn rate prediction is a critical task. Classification models are often used for churn rate prediction. But most of these models have several shortcomings.  ...  The work presented in this paper uses autoencoder network with in built attention mechanism. The Encoder part of the network compresses the data into a latent space with the help of the decoder part.  ... 
doi:10.30534/ijatcse/2019/60832019 fatcat:q26d74jtl5g6rbio3mt4hhaqvm
« Previous Showing results 1 — 15 out of 2,280 results