WITT: A Wireless Image Transmission Transformer for Semantic Communications
WITT: трансформер беспроводной передачи изображений для семантических коммуникаций
2023-05-05
SCID: 54.1/qsk8ng7w
Discuss with AI
Swin Transformerchannel state informationsemantic communicationsspatial modulationwireless image transmission
Figures from the paper
Abstract (AI)
In this paper, we aim to redesign the vision Transformer (ViT) as a new backbone to realize semantic image transmission, termed wireless image transmission transformer (WITT). Previous works build upon convolutional neural networks (CNNs), which are inefficient in capturing global dependencies, resulting in degraded end-to-end transmission performance especially for high-resolution images. To tackle this, the proposed WITT employs Swin Transformers as a more capable backbone to extract long-range information. Different from ViTs in image classification tasks, WITT is highly optimized for image transmission while considering the effect of the wireless channel. Specifically, we propose a spatial modulation module to scale the latent representations according to channel state information, which enhances the ability of a single model to deal with various channel conditions. As a result, extensive experiments verify that our WITT attains better performance for different image resolutions, distortion metrics, and channel conditions. The code is available at https://github.com/KeYang8/WITT.
Key Findings
1
A spatial modulation module scales latent representations according to channel state information, enabling one model to handle diverse channel conditions.
2
Experiments show that WITT outperforms prior approaches across different image resolutions, distortion metrics, and channel conditions.
3
The model is specifically optimized for wireless image transmission, incorporating wireless channel effects rather than directly adapting classification-oriented ViTs.
4
WITT redesigns vision Transformers for semantic image transmission, using Swin Transformers to capture long-range dependencies more effectively than CNN-based approaches.
Research Object
semantic image transmission over wireless channels using high-resolution images
Research Subject
end-to-end transmission performance and robustness across image resolutions, distortion metrics, and varying channel conditions, including channel-state-dependent latent representation scaling
Publication Details
Publication Date
2023-05-05
Journal
Publisher
ISSN
Cited by
243
Open access PDF
Access Type
Author Information
Download PDF
Subscribe to digest
References available in scid.ai4
Swin Transformer: Hierarchical Vision Transformer using Shifted Windows2021
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale2020
Deep Learning Enabled Semantic Communication Systems2021
Toward Wisdom-Evolutionary and Primitive-Concise 6G: A New Paradigm of Semantic Communication Networks2021