INNSE: Invertible Neural Network-Based DNA Image Storage with Self-Correction Encoding

Pytorch Implementation of the paper "INNSE: Invertible Neural Network-Based DNA Image Storage with Self-Correction Encoding".

Dependencies and Installation

Python 3 via Anaconda (recommended)
PyTorch >= 1.0
NVIDIA GPU + CUDA
Python Package: pip install numpy opencv-python lmdb pyyaml

Dataset Preparation

The training and test data sets for the image can be downloaded here. The training and test data sets for the video can be downloaded here.

Usage

All the implementation is in /codes. To run the code, select the corresponding configuration file in /codes/options/ and run as following command:

Training

python train.py -opt options/train/train_I_IRN_x4.yml
python train.py -opt options/train/train_V_IRN_x4.yml

Testing

python test.py -opt options/test/test_I_IRN_x4.yml
python test.py -opt options/test/test_V_IRN_x4.yml

Codec

python encode.py -opt options/en_decode/encode_I_IRN_x4.yml
python decode.py -opt options/en_decode/decode_I_IRN_x4.yml
python encode.py -opt options/en_decode/encode_V_IRN_x4.yml
python decode.py -opt options/en_decode/decode_V_IRN_x4.yml

Acknowledgement

Our project is heavily based on Invertible-Image-Rescaling and Video Rescaling Networks with Joint Optimization Strategies for Downscaling and Upscaling as basic framework.

Larissa-11/INNSE