# On Leveraging Pretrained GANs for Generation with Limited Data

arXiv: https://arxiv.org/abs/2002.11810 · HTML: https://mlsanity.com/paper/2002.11810/

## Key facts

- Cited by 11 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 26 Feb 2020.

## Built on by (most cited first, 11 of 11)

- [Few-Shot Defect Image Generation via Defect-Aware Feature Manipulation](https://mlsanity.com/paper/2303.02389/) (2023)
- [UVCGAN: UNet Vision Transformer cycle-consistent GAN for unpaired image-to-image translation](https://mlsanity.com/paper/2203.02557/) (2022)
- [Collapse by Conditioning: Training Class-conditional GANs with Limited Data](https://mlsanity.com/paper/2201.06578/) (2022)
- [When, Why, and Which Pretrained GANs Are Useful?](https://mlsanity.com/paper/2202.08937/) (2022)
- [Efficient Conditional GAN Transfer with Knowledge Propagation across Classes](https://mlsanity.com/paper/2102.06696/) (2021)
- [TransferI2I: Transfer Learning for Image-to-Image Translation from Small Datasets](https://mlsanity.com/paper/2105.06219/) (2021)
- [DigGAN: Discriminator gradIent Gap Regularization for GAN Training with Limited Data](https://mlsanity.com/paper/2211.14694/) (2022)
- [MineGAN++: Mining Generative Models for Efficient Knowledge Transfer to Limited Data Domains](https://mlsanity.com/paper/2104.13742/) (2021)
- [Transferring Unconditional to Conditional GANs with Hyper-Modulation](https://mlsanity.com/paper/2112.02219/) (2021)
- [Semi-supervised domain adaptation with CycleGAN guided by a downstream task loss](https://mlsanity.com/paper/2208.08815/) (2022)
- [Large-Scale Public Data Improves Differentially Private Image Generation Quality](https://mlsanity.com/paper/2309.00008/) (2023)

## Abstract (arXiv)

Recent work has shown generative adversarial networks (GANs) can generate highly realistic images, that are often indistinguishable (by humans) from real images. Most images so generated are not contained in the training dataset, suggesting potential for augmenting training sets with GAN-generated data. While this scenario is of particular relevance when there are limited data available, there is still the issue of training the GAN itself based on that limited data. To facilitate this, we leverage existing GAN models pretrained on large-scale datasets (like ImageNet) to introduce additional knowledge (which may not exist within the limited data), following the concept of transfer learning. Demonstrated by natural-image generation, we reveal that low-level filters (those close to observations) of both the generator and discriminator of pretrained GANs can be transferred to facilitate generation in a perceptually-distinct target domain with limited training data. To further adapt the transferred filters to the target domain, we propose adaptive filter modulation (AdaFM). An extensive set of experiments is presented to demonstrate the effectiveness of the proposed techniques on generation with limited data.