Globally and Locally Consistent Image Completion

This is a Tensorflow2 Keras implementation of "Globally and Locally Consistent Image Completion".

Install python libraries

pip install pipenv
cd image_completion_keras
pipenv install

Prepare dataset

Please download appropriate image dataset and put it under the "data" folder and "test" folder. For example, please put the images as follows.

data
└── place365
    ├── Places365_00000001.jpg
    ├── Places365_00000002.jpg
    ├── Places365_00000003.jpg
    ├── Places365_00000004.jpg
    ├── Places365_00000005.jpg
    ├── Places365_00000006.jpg
    ├── Places365_00000007.jpg
    ├── Places365_00000008.jpg
    ...
test
└── place365_test
    ├── Places365_test_00000001.jpg
    ├── Places365_test_00000002.jpg
    ├── Places365_test_00000003.jpg
    ├── Places365_test_00000004.jpg
    ├── Places365_test_00000005.jpg
    ├── Places365_test_00000006.jpg
    ├── Places365_test_00000007.jpg
    ├── Places365_test_00000008.jpg
    ...

Train

pipenv shell
python train.py

Test

pipenv shell
python test.py

Result