[2011.08382] Learning Efficient GANs via Differentiable Masks and co-Attention Distillationopen searchopen navigation menucontact arXivsubscribe to arXiv mailings

Generative Adversarial Networks (GANs) have been widely-used in image translation, but their high computational and storage costs impede the deployment on mobile devices. Prevalent methods for CNN compression cannot be directly applied to GANs due to the complicated generator architecture and the unstable adversarial training. To solve these, in this paper, we introduce a novel GAN compression method, termed DMAD, by proposing a Differentiable Mask and a co-Attention Distillation. The former searches for a light-weight generator architecture in a training-adaptive manner. To overcome channel inconsistency when pruning the residual connections, an adaptive cross-block group sparsity is further incorporated. The latter simultaneously distills informative attention maps from both the generator and discriminator of a pre-trained model to the searched generator, effectively stabilizing the adversarial training of our light-weight model. Experiments show that DMAD can reduce the Multiply Acc

3 mentions: @ak92501@maguroIsland@patri_vaquero_
Keywords: attention
Date: 2020/11/18 06:53

Referring Tweets

@maguroIsland Learning Efficient GANs via Differentiable Masks and co-Attention Distillation (CVPR2021 under review?) t.co/Sm6LyQP6is 学習できるchannel pruningを行うdifferentiable mask、generatorとdiscriminator両者のfeature mapの差分を取る項を合わせたco-Attentionを提案導入 しかし、 t.co/el9ihHoIRN
@ak92501 Learning Efficient GANs using Differentiable Masks and Co-Attention Distillation pdf: t.co/3VTb87um91 abs: t.co/c9AGrkrGn0 github: t.co/b7LFa6Tfz7 t.co/gpgUGlEUJw
@patri_vaquero_ 🔬 Learning Efficient GANs using Differentiable Masks and Co-Attention Distillation Github: t.co/Ncmr8Us2bJ Paper: t.co/R8rBqmmR8h

Related Entries

Read more LaneKeeping-Adv-Attack
0 users, 3 mentions 2020/03/05 02:20
Read more [2006.10963] Evaluating Prediction-Time Batch Normalization for Robustness under Covariate Shiftopen...
0 users, 4 mentions 2020/06/25 21:51
Read more [2007.08513] RetrieveGAN: Image Synthesis via Differentiable Patch Retrievalopen searchopen navigati...
0 users, 2 mentions 2020/07/17 03:52
Read more Layered Neural Rendering for Retiming People in Video
0 users, 3 mentions 2020/09/17 14:22
Read more GitHub - clovaai/length-adaptive-transformer
0 users, 3 mentions 2020/10/15 02:22