WatermarkRemover-AI: An Open-Source AI Watermark Removal Tool 191

What is WatermarkRemover-ai?

WatermarkRemover-AI is an open-source, practical AI watermark-removal tool. It uses Microsoft's Florence-2 technology to detect watermarks and the LaMA model to perform inpainting, effectively removing all kinds of watermarks. It supports both single-image processing and batch operations, delivers natural-looking repair results, and supports multiple output formats. It runs smoothly without requiring a GPU, and also provides a one-click installation script along with a simple and easy-to-use interface.

WatermarkRemover-AI Main Features

Dual-mode operation:

• Supports processing a single image or images in an entire directory.

Advanced watermark detection:

• Uses Florence-2's open-vocabulary detection capability to accurately identify watermarks.

Seamless inpainting:

• Uses the LaMA model for context-aware inpainting, naturally filling in the areas where watermarks have been removed.

Customizable output:

• The maximum bounding box size for watermark detection can be configured.

• The transparency of the watermark area can be set.

• Output can be forced into a specific format (PNG, WEBP, JPG).

Progress tracking:

• Provides real-time progress updates in both GUI and CLI modes.

Dark mode support:

• The GUI automatically adapts to dark mode based on system settings.

Efficient resource management:

• Supports GPU acceleration using CUDA (optional).

WatermarkRemover-AI Technical Architecture

• Florence-2: Used for watermark detection, identifying watermarks through open-vocabulary object detection.

• LaMA: Used for inpainting, context-aware filling of areas where watermarks have been removed.

• PyQt6: Used to build a user-friendly graphical interface.

How to Install WatermarkRemover-AI

Clone the repository:

git clone https://github.com/D-Ogi/WatermarkRemover-AI.git
cd WatermarkRemover-AI

Run the installation script:

• Use the setup.sh script to automatically set up the environment, install dependencies, and launch the GUI application.

• The CLI mode can also be used directly via the command line.

Manually activating the environment:

If you need to manually activate the environment, you can run:

conda activate py312aiwatermark

How to Use WatermarkRemover-AI

Graphical Interface (GUI):

• After launching the GUI, you can choose to process a single image or an entire directory.

• You can set the input/output paths and configure options (such as transparency, bounding box size, etc.).

• Click the "Start" button to begin processing.

Command Line Interface (CLI):

• Basic command format:

python remwm.py input_path output_path [options]

• Common options:

• --overwrite: Overwrite existing files.

• --transparent: Set the watermark area to transparent.

• --max-bbox-percent: Set the maximum bounding box size for watermark detection (default is 10%).

• --force-format: Force the output to a specific format (PNG, WEBP, or JPG).

Example command

python remwm.py ./input_images ./output_images --overwrite --transparent --max-bbox-percent=15 --force-format=PNG

github: https://github.com/D-Ogi/WatermarkRemover-AI