Sampling Autoencoder

Resolution independent image representation on neural network.

This implementation is based on VAE. The decoder model represents a function f(u, v, z_1, ..., z_n) which outputs the value at point (u, v) on the image.

Run

Tested on Python 3.6.3 + Chainer 3.4.0. You may need CUDA GPU.

python3 train.py --gpu 0

Output

1st epoch

Original image.

Reconstructed image rendered as 168x168.

50th epoch

Original image.

Reconstructed image rendered as 168x168 on 50th epoch.