Reproduce “Text-Guided Token Communication for Wireless Image Transmission” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Bole Liu, Li Qiao, Ye Wang, Zhen Gao, Yu Ma, Keke Ying, Tong Qin
Researchers from Beijing Institute of Technology introduced a text-guided token communication system for wireless image transmission, leveraging pre-trained multi-modal foundation models. This system delivered superior perceptual quality (LPIPS) and semantic preservation (CLIP similarity) at a 1/96 bandwidth ratio for Signal-to-Noise Ratios (SNRs) above 0 dB, compared to an ADJSCC baseline, while also mitigating reconstruction degradation at low SNRs.
Key insights
- Text guidance, particularly longer, more informative captions, significantly enhances image reconstruction quality at low SNRs and helps mitigate the "cliff effect" by enabling robust semantic restoration.
- The system's reliance on pre-trained generative foundation models allows for superior generalization to unseen datasets and diverse channel scenarios without requiring scenario-specific retraining.
- An asymmetric contribution pattern between image and text tokens exists, where image token fidelity is the primary determinant of visual quality, while text provides supplementary semantic context, enhancing cross-modal resilience.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.