/glann

Official code for paper "Non-Adversarial Image Synthesis with Generative Latent Nearest Neighbors"

Primary LanguagePythonBSD 3-Clause "New" or "Revised" LicenseBSD-3-Clause

GLANN

Official code for paper "Non-Adversarial Image Synthesis with Generative Latent Nearest Neighbors" by Yedid Hoshen, Ke Li and Jitendra Malik, CVPR'19

This repository contains Python3 implementations of:

  • Generative Latent Optimization (GLO)
  • Implicit Maximum Likelihood Estimation (IMLE)
  • Generative Latent Nearest Neighbors (GLANN)

Quick Start

Install dependencies:

pip install numpy scipy pytorch torchvision python-mnist fbpca faiss

Edit prepare_mnist.py with the correct path to the data.

Prepare a dataset:

python prepare_mnist.py

Train GLO on a particular config:

python train_glo.py configs/mnist.yaml

Train IMLE based on the trained GLO model:

python train_icp.py configs/mnist.yaml

Evaluate the FID of the trained GLANN model:

python test_fid.py configs/mnist.yaml

References

Please cite [1] if you found the resources in this repository useful.

Non-Adversarial Image Synthesis with Generative Latent Nearest Neighbors

[1] Y. Hoshen, K. Li, J. Malik, Non-Adversarial Image Synthesis with Generative Latent Nearest Neighbors

@inproceedings{hoshen2019non,
  title={Non-Adversarial Image Synthesis with Generative Latent Nearest Neighbors},
  author={Hoshen, Yedid and Li, Ke and Malik, Jitendra},
  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
  pages={5811--5819},
  year={2019}
}

Related work