This is the implementation of the paper Localization of Deep Inpainting Using High-Pass Fully Convolutional Network (ICCV 2019).

• Python 3
• Tensorflow >= 1.10.0

First, prepare the training data so that the images are stored in "xxx/jpg*/xxx/" and the corresponding groundtruth masks are stored in "xxx/msk*/xxx/". Then, run the following command.

python3 hp_fcn.py --data_dir <path_to_the_training_dataset> --logdir <path_to_the directory_for_saving_model_and_log> --mode train

Prepare the testing data in a similar way and run the code as follows.

python3 hp_fcn.py --data_dir <path_to_the_testing_dataset> --logdir <path_to_the directory_where_the_trained_model_is_saved> --mode test

The pretrained checkpoint of High-pass FCN is available at: https://drive.google.com/drive/folders/1W1f_piFIiK6JJLIXimr1vtRs8MVYLwjZ?usp=sharing

This repo also includes an implementation of MFCN (mfcn.py):

Ronald Salloum, Yuzhuo Ren, and C.-C. Jay Kuo. Image splicing localization using a multi-task fully convolutional network (MFCN). Journal of Visual Communication and Image Representation, 51:201–209, 2018.

To run the code of MFCN, you should also have the edge masks in "xxx/edg*/xxx/".

If you use our code please cite:

@InProceedings{Li_2019_ICCV,
    author = {Li, Haodong and Huang, Jiwu},
    title = {Localization of Deep Inpainting Using High-Pass Fully Convolutional Network},
    booktitle = {The IEEE International Conference on Computer Vision (ICCV)},
    pages={8301--8310},
    month = {October},
    year = {2019}
}

If you have any questions, please contact: lihaodong[AT]szu.edu.cn