1,017 Hits in 7.3 sec

Real-time Universal Style Transfer on High-resolution Images via Zero-channel Pruning [article]

Jie An, Tao Li, Haozhi Huang, Li Shen, Xuan Wang, Yongyi Tang, Jinwen Ma, Wei Liu, Jiebo Luo
2020 arXiv   pre-print
The comprehensive experiments demonstrate that ArtNet can achieve universal, real-time, and high-quality style transfer on high-resolution images simultaneously, (68.03 FPS on 512 times 512 images).  ...  Most existing algorithms use VGG19 as the feature extractor, which incurs a high computational cost and impedes real-time style transfer on high-resolution images.  ...  Via zero-channel pruning, GoogLeNet can achieve real-time efficiency in style transfer on high-resolution images.  ... 
arXiv:2006.09029v2 fatcat:n7qjeqyc2neh3mjlckgypj47ki

ThiNet: Pruning CNN Filters for a Thinner Net

Jian-Hao Luo, Hao Zhang, Hong-Yu Zhou, Chen-Wei Xie, Jianxin Wu, Weiyao Lin
2018 IEEE Transactions on Pattern Analysis and Machine Intelligence  
We formally establish filter pruning as an optimization problem, and reveal that we need to prune filters based on statistics computed from its next layer, not the current layer, which differentiates ThiNet  ...  We focus on filter level pruning, i.e., the whole filter will be discarded if it is less important. An effective and unified framework, ThiNet (stands for "Thin Net"), is proposed in this paper.  ...  We use a recently proposed arbitrary style transfer algorithm AdaIN-Style [6] in our experiment. Since the VGG19 network [2] is more effective in style transfer, we pruned a VGG19 network first.  ... 
doi:10.1109/tpami.2018.2858232 pmid:30040622 fatcat:m2kdsmx2enan7fjqzhjfrs2gve

Learning Slimming SAR Ship Object Detector through Network Pruning and Knowledge Distillation

Shiqi Chen, Ronghui Zhan, Wei Wang, Jun Zhang
2020 IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing  
The deployment of deep convolutional neural networks (CNNs) in synthetic aperture radar (SAR) ship real-time detection is largely hindered by huge computational cost.  ...  This is achieved by self-designed backbone structure and network pruning, which enforces channel-level sparsity in the backbone network and yields a compact model.  ...  It comprises 31 single-polarized and high-resolution SAR images acquired from GF-3 satellite.  ... 
doi:10.1109/jstars.2020.3041783 fatcat:e4cnohcfq5eu5nsg6ke6feidoi

Real-time Localized Photorealistic Video Style Transfer [article]

Xide Xia, Tianfan Xue, Wei-sheng Lai, Zheng Sun, Abby Chang, Brian Kulis, Jiawen Chen
2020 arXiv   pre-print
Our method, based on a deep neural network architecture inspired by recent work in photorealistic style transfer, is real-time and works on arbitrary inputs without runtime optimization once trained on  ...  We demonstrate our method on a variety of style images and target videos, including the ability to transfer different styles onto multiple objects simultaneously, and smoothly transition between styles  ...  network for photorealistic video style transfer that runs in real-time (26.5 Hz at 1024 × 1024 resolution).  ... 
arXiv:2010.10056v1 fatcat:n2pv3dtg3zfifnnj6tm3mge6w4

Silicon photonic subspace neural chip for hardware-efficient deep learning [article]

Chenghao Feng, Jiaqi Gu, Hanqing Zhu, Zhoufeng Ying, Zheng Zhao, David Z. Pan, Ray T. Chen
2021 arXiv   pre-print
We experimentally demonstrate our PSNN on a butterfly-style programmable silicon photonic integrated circuit and show its utility in practical image recognition tasks.  ...  The optical neural network (ONN) is a promising candidate for next-generation neurocomputing due to its high parallelism, low latency, and low energy consumption.  ...  We experimentally demonstrate our PSNN on a butterfly-style programmable silicon photonic integrated circuit and show its utility in practical image recognition tasks.  ... 
arXiv:2111.06705v1 fatcat:5yp4yfg6jjbtpib2nvrd3dhoyq

FreezeNet: Full Performance by Reduced Storage Costs [article]

Paul Wimmer, Jens Mehnert, Alexandru Condurache
2020 arXiv   pre-print
Pruning generates sparse networks by setting parameters to zero.  ...  On the classification tasks MNIST and CIFAR-10/100 we outperform SNIP, in this setting the best reported one-shot pruning method, applied before training.  ...  .: Learning multiple layers of features from tiny images. University of Toronto (2012), 7 27.  ... 
arXiv:2011.14087v1 fatcat:mazpvgnxxnaw7psj2gev7ovhde

Dynamic Slimmable Network [article]

Changlin Li, Guangrun Wang, Bing Wang, Xiaodan Liang, Zhihui Li, Xiaojun Chang
2021 arXiv   pre-print
However, dynamic sparse patterns on convolutional filters fail to achieve actual acceleration in real-world implementation, due to the extra burden of indexing, weight-copying, or zero-masking.  ...  Typically, DS-Net achieves 2-4x computation reduction and 1.62x real-world acceleration over ResNet-50 and MobileNet with minimal accuracy drops on ImageNet.  ...  We also test the transferability of our DS-Net on CIFAR-10, which comprises 10 classes with 50,000 training and 10,000 test images.  ... 
arXiv:2103.13258v1 fatcat:q3dqpaakf5bilikvsn7akzhone

TimbreTron: A WaveNet(CycleGAN(CQT(Audio))) Pipeline for Musical Timbre Transfer [article]

Sicong Huang, Qiyang Li, Cem Anil, Xuchan Bao, Sageev Oore, Roger B. Grosse
2019 arXiv   pre-print
We introduce TimbreTron, a method for musical timbre transfer which applies "image" domain style transfer to a time-frequency representation of the audio signal, and then produces a high-quality waveform  ...  In principle, one could apply image-based style transfer techniques to a time-frequency representation of an audio signal, but this depends on having a representation that allows independent manipulation  ...  An appealing strategy would be to directly apply image-based style transfer techniques to time-frequency representations of images, such as short-time Fourier transform (STFT) spectrograms.  ... 
arXiv:1811.09620v2 fatcat:sk6fqvmdzna4hnouhwalpjhzxu

Name Your Style: An Arbitrary Artist-aware Image Style Transfer [article]

Zhi-Song Liu, Li-Wen Wang, Wan-Chi Siu, Vicky Kalogeiton
2022 arXiv   pre-print
Moreover, it can mimic the styles of one or many artists to achieve attractive results, thus highlighting a promising direction in image style transfer.  ...  In this paper, we propose a text-driven image style transfer (TxST) that leverages advanced image-text encoders to control arbitrary style transfer.  ...  To handle high-resolution images, ArtNet [13] proposes Zero-channel Pruning to reduce model complexity. Wang et al.  ... 
arXiv:2202.13562v2 fatcat:vbhnuyar2fc2ncsfakip42enry

Neural Network Model for Detection of Edges Defined by Image Dynamics

Patrick A. Shoemaker
2019 Frontiers in Computational Neuroscience  
visual field, based on general differences in image dynamics.  ...  Insects can detect the presence of discrete objects in their visual fields based on a range of differences in spatiotemporal characteristics between the images of object and background.  ...  (B) Shows the central part of the visual array, with the photoreceptors stimulated by a low-resolution version (red) of a high-resolution, time-varying image (blue).  ... 
doi:10.3389/fncom.2019.00076 pmid:31787888 pmcid:PMC6854273 fatcat:immwqzlhkfg4bldk4frdp4tseu

PEPSI++: Fast and Lightweight Network for Image Inpainting [article]

Yong-Goo Shin, Min-Cheol Sagong, Yoon-Jae Yeo, Seung-Wook Kim, Sung-Jea Ko
2020 arXiv   pre-print
Simultaneously, the inpainting path generates higher inpainting quality using the refined features reconstructed via the CAM.  ...  (PSNR) and structural similarity (SSIM), as well as significantly reduce hardware costs such as computational time and the number of network parameters.  ...  At the same time, the inpainting path produces image with high quality using the refined features reconstructed via the CAM.  ... 
arXiv:1905.09010v4 fatcat:j562nzwoprhedipkghfhq7so7y

A Survey on Generative Adversarial Networks: Variants, Applications, and Training [article]

Abdul Jabbar, Xi Li, Bourahla Omar
2020 arXiv   pre-print
The Generative Models have gained considerable attention in the field of unsupervised learning via a new and practical framework called Generative Adversarial Networks (GAN) due to its outstanding data  ...  Image super-resolution The purpose of image super-resolution (SR) is to up-scale the images of low-resolution (LR) because images with high-resolution (HR) have better visual qualities.  ...  Facial makeup transfer In facial makeup transfer, the facial makeup style of one face is transfer to a non-makeup face without losing the identity of the non-makeup face, and makeup removal system does  ... 
arXiv:2006.05132v1 fatcat:gyjezuh5sfdilkp43ydsea5cwa

Table of contents

2021 ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)  
Serra, Emilia Gómez, Universitat Pompeu Fabra, Spain MLSP-23.6: MICAUGMENT: ONE-SHOT MICROPHONE STYLE TRANSFER ................................................ 3080 Zalán Borsos, ETH Zurich, Switzerland  ...  of China, China; Tie-Yan Liu, Microsoft Research Asia, China SPE-3.6: A NEW HIGH QUALITY TRAJECTORY TILING BASED HYBRID TTS IN ........................................ 5704 REAL TIME Feng-Long Xie, Xin-Hui  ... 
doi:10.1109/icassp39728.2021.9414617 fatcat:m5ugnnuk7nacbd6jr6gv2lsfby

Table of Contents

2019 2019 IEEE/CVF International Conference on Computer Vision (ICCV)  
Jiao Tong University of China) xxxiv Oral 2.1B Low Level Vision Wavelet Domain Style Transfer for an Effective Perception-Distortion Tradeoff in Single Image Super-Resolution 3076 Xin Deng  ...  CG2Real: Synthetic-to-Real Translation via Image Disentanglement 2730 Sai Bi (University of California.  ... 
doi:10.1109/iccv.2019.00004 fatcat:5aouo4scprc75c7zetsimylj2y

Table of Contents

2021 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)  
A Two-Stage Deep Network for High Dynamic Range Image Reconstruction 550  ...  organizers), Kim Byeoung-su (Mobile AI 2021 challenge organizers), Radu Timofte (Mobile AI 2021 challenge organizers), and Angeline Pouget (missing) Real-Time Quantized Image Super-Resolution on Mobile  ...  Skinner (University of Michigan), Travis Driver (Georgia Institute of Technology), and Panagiotis Tsiotras (Georgia Institute of Technology) Improving Astronomy Image Quality through Real-Time Wavefront  ... 
doi:10.1109/cvprw53098.2021.00004 fatcat:yh3zw4afzneeza6jrd6377fd3y
« Previous Showing results 1 — 15 out of 1,017 results