Frequency-Guided Iterative Network for Image Steganography (FIS)

This is the official code for the manuscript [Enhancing Image Steganography via Frequency-Guided Iterative Optimization]. [I don't know why, after having AI standardize the code, it seems like the reasoning is a bit off? 不知道为什么让AI规范了一下代码后好像推理有点问题?]

Getting Started

Prerequisites

Environment

• Python >= 3.6
• PyTorch >= 1.10.2

Datasets

Download the datasets subsampled image datasets (same as LISO). It is the subset of Div2k, CelebA, and MS COCO. To use them, set your dataset path in the 'utils.py'.

Training

python main.py --bits 2 --dataset div2k

Evaluation

python main.py --eval --bits 2 --dataset div2k --load checkpoints/div2k/2_bits.steg

Running Example

This is the running example of my project.

Acknowledgements

• LISO
• CBAM
• TrajGRU
• Extensive Reading During Research