Deep-Image-Analogy

This project is a python implementation of Deep Image Analogy.

Some results

Requirements

• python3

• opencv3

  If you use anaconda, you can install opencv3 by conda install opencv

• pytorch (>= 1.7 for cpu-only mode, >= 1.3 for gpu mode)

• pycuda (optional)

Usage (demo)

Use GPU

python main.py --use_cuda

CPU only

python main.py

I train a series of invert blocks which estimate the inputs given the outputs for each stage of Vgg19. The results are slightly worse than optimization-based method.

Acknowledgments

My project acknowledge the official code Deep-Image-Analogy, pytorch, and another pytorch implementation. Especially, thanks for the authors of this amazing algorithm.