not-so-BigGAN: Generating High-Fidelity Images on a Small Compute Budget

By Seung-Jun Han, Akash Srivastava, C. Hurwitz, Prasanna Sattigeri, David D. Cox. 2020

10 citations recorded by the source corpus.

BigGAN is the state-of-the-art in high-resolution image generation, successfully leveraging advancements in scalable computing and theoretical understanding of generative adversarial methods to set new records in conditional image generation. A major part of BigGAN's success is due to its use of large mini-batch sizes during training in high dimensions. While effective, this technique requires an incredible amount of compute resources and/or time (256 TPU-v3 Cores), putting the model out of reach for the larger research community. In this paper, we present not-so-BigGAN, a simple and scalable framework for training deep generative models on high-dimensional natural images. Instead of modelling the image in pixel space like in BigGAN, not-so-BigGAN uses wavelet transformations to bypass the curse of dimensionality, reducing the overall compute requirement significantly. Through extensive empirical evaluation, we demonstrate that for a fixed compute budget, not-so-BigGAN converges several times faster than BigGAN, reaching competitive image quality with an order of magnitude lower compute budget (4 Telsa-V100 GPUs).

Subfields: Computer Vision and Pattern Recognition, Generative Models, Efficiency & Compression

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

1. Bo Dai
2. Chen Change Loy
3. Wayne Wu
4. Liming Jiang
5. Luc Van Gool
6. Xiaogang Wang
7. Ming-Hsuan Yang
8. Dacheng Tao

In Generative Models

1. Bo Dai
2. Chen Change Loy
3. Wayne Wu
4. Liming Jiang
5. William L. Hamilton
6. Le Song
7. Tommi S. Jaakkola

In Efficiency & Compression

1. Bo Dai
2. Chen Change Loy
3. Chao Dong

Citations in the corpus

• Focal Frequency Loss for Image Reconstruction and Synthesis (2020, background): “Besides, several studies have incorporated frequency analysis with network compression [16, 66, 68, 6, 17] and feature reduction [35, 61] to accelerate the training and inference of networks.”