zuruoke/watermark-removal

Note: The Google Colab method is currently marked as broken in the repository. If attempting to use it, follow these steps:

1. Clone the repository: !git clone https://github.com/zuruoke/watermark-removal
2. Change directory to the repo.
3. Downgrade TensorFlow to version 1.15.0: !pip install tensorflow==1.15.0
4. Install the neuralgym toolkit: !pip install git+https://github.com/JiahuiYu/neuralgym
5. Download the model directory and place it in model/. Ensure checkpoint.txt is renamed to checkpoint.
6. Run the removal using main.py.

!python main.py --image path-to-input-image --output path-to-output-image --checkpoint_dir model/ --watermark_type istock

You can run the watermark removal process inside a Docker container. This requires building the image from the repository and providing the pre-trained model directory via a volume mount.

1. Clone the repository.
2. Build the Docker image: docker build -t watermark-removal .
3. Download the model directory from this link.
4. Run the container by mounting the model, input, and output directories.

docker run --rm -v '<path_to_model_dir>:/repo/model' -v '<path_to_input_dir>:/input' -v '<path_to_output_dir>:/output' watermark-removal --checkpoint_dir /repo/model --image '/input/<input_image_file>' --output '/output/<output_image_file>' --watermark_type istock

When running the tool (either via Docker or directly via main.py), use the following arguments:

• --checkpoint_dir: Path to the directory containing the model checkpoints.
• --image: Path to the input image file containing the watermark.
• --output: Path where the processed image should be saved.
• --watermark_type: The type of watermark to remove (e.g., istock).

The main.py script provides a command-line interface to remove watermarks from images using a pre-trained inpainting model. It requires an input image, an output path, the type of watermark to target, and the directory containing the TensorFlow model checkpoints.

Arguments:

• --image: The filename of the image to be processed (string).
• --output: The path where the processed image will be saved (string, default: output.png).
• --watermark_type: The specific type of watermark to remove (string, default: istock).
• --checkpoint_dir: The directory containing the TensorFlow checkpoints (string, default: model/).