brownbear2polarbear

Experiment using CycleGAN for Image to Image translation on images of brown bears and polar bears.

Please note that the model does NOT create accurate output, this is most likely due to the fact that the background of the images in A and B are not similiar.

Setup for training

1. Create a folder in your google drive named Colab Notebooks (if it does not already exist)
2. Create a folder inside Colab Notebooks called brownbear2polarbear
3. Create a .zip archive called "trainA" containing a folder called "trainA" containing images of brown bears you would like to train the model on.
4. Create a .zip archive called "trainB" containing a folder called "trainB" containing images of polar bears you would like to train the model on.
5. Place this .zip archive inside the folder brownbear2polarbear that we created in step 2.
6. Copy the folder brownbear2polarbear_checkpoint to your google drive (inside Colab Notebooks, not brownbear2polarbear).
7. Now you should be ready to open brownbear2polarbear_cyclegan.ipynb in CoLab and run it, click the badge on the top of this page to open this in colab.
8. After running, the results should be stored in brownbear2polarbear_results, along with a copy in a folder labeled after the date and time of completing the run.

Setup for testing

1. Create a folder in your google drive named Colab Notebooks (if it does not already exist)
2. Create a folder inside Colab Notebooks called brownbear2polarbear
3. Create a .zip archive called "testA" containing a folder called "testA" containing images of brown bears you would like to transform.
4. Create a .zip archive called "testB" containing a folder called "testB" containing images of polar bears you would like to transform.
5. Place this .zip archive inside the folder brownbear2polarbear that we created in step 2.
6. Copy the folder brownbear2polarbear_checkpoint to your google drive (inside Colab Notebooks, not brownbear2polarbear).
7. Now you should be ready to open brownbear2polarbear_cyclegan_only_run_tests.ipynb in CoLab and run it, click the badge on the top of this page to open this in colab.
8. After running, the results should be stored in brownbear2polarbear_results, along with a copy in a folder labeled after the date and time of completing the run.

Copyright

All images are taken with permission for non-commercial and educational use from ImageNet.