Deep-Learning-Enabled-Semantic-Communication-Systems-with-Task-Unaware-Transmitter-and-Dynamic-Data

pytorch implementation of "Deep Learning-Enabled Semantic Communication Systems with Task-Unaware Transmitter and Dynamic Data"

Prerequites

• [Python 3.7]
• [PyTorch 0.1.12]
• [Torchvision 0.9.1]
• [Torch 1.5.1]
• [Numpy 1.21.2]

The folders

"semantic_extraction" and "semantic_system_with_DA" are the semantic extraction part and the data adaptation part of the proposed method in the paper.

The details are represented in the two sub-folders.

Other Instructions

This is an example of semantic communication using a small-sized dataset based on MLP and CNN. If you require a more advanced neural network framework or a system with better performance, we recommend using our another code repository based on Swin Transformer

Citation

Please use the following BibTeX citation if you use this repository in your work:

@ARTICLE{9953099,
  author={Zhang, Hongwei and Shao, Shuo and Tao, Meixia and Bi, Xiaoyan and Letaief, Khaled B.},
  journal={IEEE Journal on Selected Areas in Communications}, 
  title={Deep Learning-Enabled Semantic Communication Systems With Task-Unaware Transmitter and Dynamic Data}, 
  year={2023},
  volume={41},
  number={1},
  pages={170-185},
  doi={10.1109/JSAC.2022.3221991}}