Asymmetric Semantic Communication System Based on Diffusion Model in IoT
Tianjing Ren, Hao Wu · 2023
This paper presents an asymmetric semantic communication network based on the diffusion model for transmitting and recovering semantic information from images. The network employs a semantic segmentation model to extract semantic information from the images, generate desired images using the diffusion model. We compared end-to-end semantic communication networks with GAN-based semantic communication networks. The results demonstrate that the asymmetric semantic communication network based on the diffusion model outperforms in terms of accuracy, PSNR, FID, and exhibits good robustness. The asymmetric semantic communication network offers efficient and high-performance image transmission in situations where computational resources are limited on IoT devices.