Need help with chainer-pix2pix?
Click the “chat” button below for chat support from the developer who created it, or find similar developers for support.

About the developer

pfnet-research
127 Stars 39 Forks 16 Commits 4 Opened issues

Description

chainer implementation of pix2pix

Services available

!
?

Need anything else?

Contributors list

chainer-pix2pix

chainer implementation of pix2pix https://phillipi.github.io/pix2pix/

The Japanese readme can be found here.

Example result on CMP facade dataset

From the left side: input, output, ground_truth

usage

  1. pip install -r requirements.txt
  2. Download the facade dataset (base set) http://cmp.felk.cvut.cz/~tylecr1/facade/
  3. python train_facade.py -g [GPU ID, e.g. 0] -i [dataset root directory] --out [output directory] --snapshot_interval 10000
  4. Wait a few hours...
    • --out
      stores snapshots of the model and example images at an interval defined by
      --snapshot_interval
    • If the model size is large, you can reduce
      --snapshot_interval
      to save resources.

Using other datasets

  • Gather image pairs (e.g. label + photo). Several hundred pairs are required for good results.
  • Create a copy of
    facade_dataset.py
    for your dataset. The function get_example should be written so that it returns the i-th image pair a tuple of numpy arrays i.e.
    (input, output)
    .
  • It maybe necessary to update the loss function in
    updater.py
    .
  • Likewise, make a copy of
    facade_visualizer.py
    and modify to visualize the dataset.
  • In
    train_facade.py
    change
    in_ch
    and
    out_ch
    to the correct input and output channels for your data.

We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.