semantic_extraction

The semantic extraction part of the proposed method in "Deep Learning-Enabled Semantic Communication Systems with Task-Unaware Transmitter and Dynamic Data".

Important Update

MLP_MNIST_model.py, MNIST.py, googlenet_train.py, and CIFAR.py have been updated on 2024.11.26.

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

Notes

The folder is for image classification task with the MNIST and CIFAR10 datasets. The image segmentation task with the PASCAL-VOC dataset is in the sub-folder VOC.

Quick Start

Train the Classifier (Pragmatic Function)

1) For the MNIST dataset

$ python MLP_MNIST_model.py 

2) For the CIFAR10 dataset

$ python googlenet_train.py 

Train the Semantic Extraction Part

1) For the MNIST dataset

$ python MNIST.py --alpha xx --pretrain_epoch xx --random_seed xx

2) For the CIFAR10 dataset

$ python CIFAR.py --alpha xx --pretrain_epoch xx --random_seed xx

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}}