mlsanityData to 14 Oct 2026

Towards Device Efficient Conditional Image Generation

arXiv 2203.10363 ↗First posted 19 Mar 2022cs.CV cs.GR eess.IVcited by 0

Not on the map. Why?

Abstract (arXiv)

We present a novel algorithm to reduce tensor compute required by a conditional image generation autoencoder without sacrificing quality of photo-realistic image generation. Our method is device agnostic, and can optimize an autoencoder for a given CPU-only, GPU compute device(s) in about normal time it takes to train an autoencoder on a generic workstation. We achieve this via a two-stage novel strategy where, first, we condense the channel weights, such that, as few as possible channels are used. Then, we prune the nearly zeroed out weight activations, and fine-tune the autoencoder. To maintain image quality, fine-tuning is done via student-teacher training, where we reuse the condensed autoencoder as the teacher. We show performance gains for various conditional image generation tasks: segmentation mask to face images, face images to cartoonization, and finally CycleGAN-based model over multiple compute devices. We perform various ablation studies to justify the claims and design choices, and achieve real-time versions of various autoencoders on CPU-only devices while maintaining image quality, thus enabling at-scale deployment of such autoencoders.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 49 of its 58 references that are in the map. First posted on arXiv 19 Mar 2022.

Builds on

49 of 58 references.

References · 49

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2014Adam: A Method for Stochastic Optimizationcited by 37,240
  3. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,305
  4. 2015U-Net: Convolutional Networks for Biomedical Image Segmentationcited by 19,713
  5. 2014Very Deep Convolutional Networks for Large-Scale Image Recognitioncited by 15,364
  6. 2015Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shiftcited by 7,005
  7. 2016Densely Connected Convolutional Networkscited by 6,720
  8. 2019EfficientNet: Rethinking Model Scaling for Convolutional Neural Networkscited by 6,442
  9. 2018MobileNetV2: Inverted Residuals and Linear Bottleneckscited by 4,795
  10. 2017Squeeze-and-Excitation Networkscited by 4,789
  11. 2016Image-to-Image Translation with Conditional Adversarial Networkscited by 4,527
  12. 2014Going Deeper with Convolutionscited by 4,241
37 more
  1. 2017MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applicationscited by 3,845
  2. 2015Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Codingcited by 2,329
  3. 2015Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networkscited by 2,054
  4. 2015Learning both Weights and Connections for Efficient Neural Networkscited by 2,031
  5. 2016Identity Mappings in Deep Residual Networkscited by 1,965
  6. 2016Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Networkcited by 1,908
  7. 2019Semantic Image Synthesis with Spatially-Adaptive Normalizationcited by 1,299
  8. 2017ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile Devicescited by 1,200
  9. 2017High-Resolution Image Synthesis and Semantic Manipulation with Conditional GANscited by 1,179
  10. 2016Pruning Filters for Efficient ConvNetscited by 1,143
  11. 2017Learning Transferable Architectures for Scalable Image Recognitioncited by 1,101
  12. 2016SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and <0.5MB model sizecited by 1,060
  13. 2017Channel Pruning for Accelerating Very Deep Neural Networkscited by 702
  14. 2017Learning Efficient Convolutional Networks through Network Slimmingcited by 676
  15. 2019MaskGAN: Towards Diverse and Interactive Facial Image Manipulationcited by 628
  16. 2017To prune, or not to prune: exploring the efficacy of pruning for model compressioncited by 603
  17. 2016Learning Structured Sparsity in Deep Neural Networkscited by 541
  18. 2017ThiNet: A Filter Level Pruning Method for Deep Neural Network Compressioncited by 442
  19. 2019GhostNet: More Features from Cheap Operationscited by 410
  20. 2018Video-to-Video Synthesiscited by 390
  21. 2018Filter Pruning via Geometric Median for Deep Convolutional Neural Networks Accelerationcited by 367
  22. 2018Model compression via distillation and quantizationcited by 289
  23. 2020StyleRig: Rigging StyleGAN for 3D Control over Portrait Imagescited by 268
  24. 2016Network Trimming: A Data-Driven Neuron Pruning Approach towards Efficient Deep Architecturescited by 221
  25. 2017SMASH: One-Shot Model Architecture Search through HyperNetworkscited by 202
  26. 2017Evolving Deep Neural Networkscited by 95
  27. 2020Practical Deep Raw Image Denoising on Mobile Devicescited by 76
  28. 2018A Style-Aware Content Loss for Real-time HD Style Transfercited by 58
  29. 2021Anycost GANs for Interactive Image Synthesis and Editingcited by 54
  30. 2020AutoGAN-Distiller: Searching to Compress Generative Adversarial Networkscited by 50
  31. 2020GAN Slimming: All-in-One GAN Compression by A Unified Optimization Frameworkcited by 44
  32. 2020Distilling portable Generative Adversarial Networks for Image Translationcited by 37
  33. 2019Co-Evolutionary Compression for Unpaired Image Translationcited by 32
  34. 2019Compressing GANs using Knowledge Distillationcited by 31
  35. 2020Slimmable Generative Adversarial Networkscited by 27
  36. 2021Revisiting Discriminator in GAN Compression: A Generator-discriminator Cooperative Compression Schemecited by 23
  37. 2017Efficient Architecture Search by Network Transformationcited by 9

Built on by

Not cited yet.