zuruoke/watermark-removal
a machine learning image inpainting task that instinctively removes watermarks from image indistinguishable from the ground truth image
What it solves
This project provides a way to remove watermarks from images using machine learning, aiming to produce results that are indistinguishable from the original, watermark-free versions of the image.
How it works
It employs a machine learning-based image inpainting methodology. The implementation is inspired by research on Contextual Attention and Gated Convolution to fill in the missing pixels where the watermark was located.
Who it’s for
Users looking to clean up images by removing unwanted watermarks using deep learning techniques.
Highlights
- Uses TensorFlow 1.15.0 for image inpainting.
- Inspired by CVPR 2018 (Contextual Attention) and ICCV 2019 (Gated Convolution) research.
- Supports Docker for easy deployment and execution.
- Provides pre-trained model checkpoints for specific watermark types (e.g., istock).
Related
- Project
- Project
- Dispatch
- Project
- Dispatch