HiDDeN
Pytorch implementation of paper "HiDDeN: Hiding Data With Deep Networks" by Jiren Zhu, Russell Kaplan, Justin Johnson, and Li Fei-Fei: https://arxiv.org/abs/1807.09937
This is an improved implementation of https://github.com/ando-khachatryan/HiDDeN. The original README file is here.
The authors have Lua+Torch implementation here: https://github.com/jirenz/HiDDeN
You need Pytorch 1.0 with TorchVision to run this.
We use 10,000 images for training and 1,000 images for validation. Following the original paper, we chose those 10,000 + 1,000 images randomly from one of the coco datasets. http://cocodataset.org/#download
The data directory has the following structure:
<data_root>/
train/
train_class/
train_image1.jpg
train_image2.jpg
...
val/
val_class/
val_image1.jpg
val_image2.jpg
...
train_class
and val_class
folders are so that we can use the standard torchvision data loaders without change.