cnn-watermark-removal

Fully convolutional deep neural network to remove transparent overlays from images

Github stars Tracking Chart

Warning! The architecture used in this project does not generalize well. You may want to check https://dmitryulyanov.github.io/deep_image_prior. This inpainting technique will likely give you better results.

Fully convolutional watermark removal attack

Deep learning architecture to remove transparent overlays from images.

example

Top: left is with watermark, middle is reconstruction and right is the mask the algo predicts (the neural net was never trained using text or this image)

Bottom: Pascal dataset image reconstructions. When the watermarked area is saturated, the reconstruction tends to produce a gray color.

Design choices

At train time, I generate a mask. It is a rectangle with randomly generated parameters (height, width, opacity, black/white, rotation). The mask is applied to a picture and the network is trained to find what was added. The loss is abs(prediction, image_perturbations)**1/2. It is not on the entire picture. An area around the mask is used to make the problem more tractable.

The network architecture does not down-sample the image. The prediction with a down-sampling network were not accurate enough. To have a large enough receptive field and not blow up the compute, I use dilated convolution. So concretely, I have a densenet style block, a bunch of dilated convolutions and final convolution to output a picture (3 channels). I did not spend much time doing hyper-parameters optimization. There's room to get better results using the current architecture.

Limitations: this architectures does not generalize to watermarks that are too different from the one generated with create_mask and it produces decent results only when the overlay is applied in an additive fashion.

Usage

This project uses Tensorflow. Install packages withpip install -r requirements.txt

You will need the jpeg library to compile Pillow from source: sudo apt-get install libjpeg-dev zlib1g-dev

You will also need to download the pascal dataset (used by default) from http://host.robots.ox.ac.uk/pascal/VOC/voc2012/ or CIFAR10 python version from https://www.cs.toronto.edu/~kriz/cifar.html (use flag --dataset=dataset_cifar). Make sure the extract the pascal dataset under a directory called data. The project directory should then have the directory cifar-10-batches-py and/or data/VOCdevkit/VOC2012/JPEGImages. If you want to use your own images, place them in data/VOCdevkit/VOC2012/JPEGImages/.

To train the network python3 watermarks.py --logdir=save/. It starts to produce some interesting results after 12000 steps.

To use the network for inference, you can run python watermarks.py --image assets/cat.png --selection assets/cat-selection.png this will create a new image output.png.

Pretrained weights

Here you can find the weights: https://github.com/marcbelmont/cnn-watermark-removal/files/1594328/data.zip
put them in /tmp/

Main metrics

Overview
Name With Ownermarcbelmont/cnn-watermark-removal
Primary LanguagePython
Program languagePython (Language Count: 1)
Platform
License:
所有者活动
Created At2017-09-30 14:30:05
Pushed At2021-03-29 17:02:59
Last Commit At2019-12-27 10:39:26
Release Count0
用户参与
Stargazers Count1.3k
Watchers Count50
Fork Count227
Commits Count25
Has Issues Enabled
Issues Count42
Issue Open Count20
Pull Requests Count5
Pull Requests Open Count2
Pull Requests Close Count2
项目设置
Has Wiki Enabled
Is Archived
Is Fork
Is Locked
Is Mirror
Is Private