This repository provides the official implementation of our IEEE JSAC 2025 paper titled:
"Multidomain Adaptive Semantic Communications"
IEEE Journal on Selected Areas in Communications (JSAC), Vol. 43, No. 7, pp. 2506–2517, 2025
DOI: 10.1109/JSAC.2025.3559127
We propose two multidomain-adaptive modules for semantic image communication over wireless channels:
-
MASCN (Multidomain Adaptive Semantic Coding Network)
A StarGAN-based end-to-end encoder-decoder architecture that inherently supports multi-domain semantic encoding and decoding, making it robust against both domain shifts and channel noise (e.g., AWGN). -
MDAN (Multidomain Data Adaptation Network)
A generalized, StarGAN-based adaptation module derived from the CycleGAN-like framework in [4], enabling multi domain data adaptation to desired target domains, prior to semantic endcoding.
Both components support multi-domain settings in a unified and scalable manner, and can be combined or used independently.
4-digit Datasets https://drive.google.com/drive/folders/1joqfCtRxitmrTZao49smzVnLT_6FSNRQ?usp=sharing
Our work is inspired by and partially built upon the following public repositories:
-
[1] DeepJSCC-f (JSCC with feedback): DeepJSCC-f: Deep Joint Source-Channel Coding of Images with Feedback Kurka & Gunduz, IEEE JSAIT 2020
👉 https://github.com/kurka/deepJSCC-feedback -
[2] ADJSCC (with attention modules): Wireless Image Transmission Using Deep Source Channel Coding With Attention Modules Xu et al., IEEE TCSVT 2022
👉 https://github.com/alexxu1988/ADJSCC -
[3] Predictive and Adaptive Deep Coding for Wireless Image Transmission in Semantic Communication Zhang et al., IEEE TWC 2023
👉 https://github.com/wyzhang-ustb/Predictive-and-Adaptive-Deep-Coding-for-Wireless-Image-Transmission-in-Semantic-Communication -
[4] Deep Learning-Enabled Semantic Communication Systems with Task-Unaware Transmitter and Dynamic Data Zhang et al., IEEE JSAC 2022
👉 https://github.com/SJTU-mxtao/Semantic-Communication-Systems
If you use this code or find the project helpful, please cite:
@ARTICLE{10960416,
author={Won, Dongwook and Do, Quang Tuan and Win, Thwe Thwe and Lee, Donghyun and Oh, Junsuk and Cho, Sungrae},
journal={IEEE Journal on Selected Areas in Communications},
title={Multidomain Adaptive Semantic Communications},
year={2025},
volume={43},
number={7},
pages={2506-2517},
keywords={Encoding;Semantic communication;Decoding;Signal to noise ratio;Adaptive systems;Transmitters;Generators;Receivers;Scalability;Adaptation models;Semantic communications;semantic coding;domain adaptation},
doi={10.1109/JSAC.2025.3559127}
}