Generating High Fidelity Data from Low-density Regions using Diffusion Models
About
Our work focuses on addressing sample deficiency from low-density regions of data manifold in common image datasets. We leverage diffusion process based generative models to synthesize novel images from low-density regions. We observe that uniform sampling from diffusion models predominantly samples from high-density regions of the data manifold. Therefore, we modify the sampling process to guide it towards low-density regions while simultaneously maintaining the fidelity of synthetic data. We rigorously demonstrate that our process successfully generates novel high fidelity samples from low-density regions. We further examine generated samples and show that the model does not memorize low-density data and indeed learns to generate novel samples from low-density regions.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Class-conditional Image Generation | ImageNet 256x256 | -- | 1021 | |
| Image Generation | ImageNet 64x64 | FID5.696 | 130 | |
| Unconditional Image Generation | CelebA 64x64 | cFID61.61 | 6 |