Fig 9 - available via license: Creative Commons Attribution 4.0 International
Content may be subject to copyright.
Download
View publication
The rate-semantic-perceptual curves of different image codecs. At a certain bitrate, lower semantic loss and lower perceptual loss means better performance.

The rate-semantic-perceptual curves of different image codecs. At a certain bitrate, lower semantic loss and lower perceptual loss means better performance.

Source publication
Fig. 2: The architecture of the RL-based Semantic Bit Allocation Model.
Fig. 3: The architecture of the semantic decoder.
Fig. 5: The semantic performance in terms of mIoU of different image...
Fig. 8: The perceptual performance of different image codecs in terms...
Fig. 9: The rate-semantic-perceptual curves of different image codecs....
Towards Semantic Communications: Deep Learning-Based Image Semantic Coding
Preprint
Full-text available
  • Aug 2022
Semantic communications has received growing interest since it can remarkably reduce the amount of data to be transmitted without missing critical information. Most existing works explore the semantic encoding and transmission for text and apply techniques in Natural Language Processing (NLP) to interpret the meaning of the text. In this paper, we...
Cite
Download full-text

Contexts in source publication

Context 1
... can further illustrate the triple trade-off rate-semantic-perceptual of the proposed RL-ASC method and the baselines in Fig. 9. In particular, the semantic loss is defined as (1-mIoU), considering the semantic segmentation task. The perceptual loss is defined as the weighted addition of FID and KID values, where a lower value means better performance. As shown in Fig. 9, the proposed RL-ASC achieves lower semantic and perceptual loss compared to the baselines ...
Context 2
... the triple trade-off rate-semantic-perceptual of the proposed RL-ASC method and the baselines in Fig. 9. In particular, the semantic loss is defined as (1-mIoU), considering the semantic segmentation task. The perceptual loss is defined as the weighted addition of FID and KID values, where a lower value means better performance. As shown in Fig. 9, the proposed RL-ASC achieves lower semantic and perceptual loss compared to the baselines at equal bitrate by a large margin, which validates the effectiveness of the proposed method in semantic exchange and visual performance. Note that the deep learning based methods RL-ASC, [3], [32] outperform the classic engineered codecs, which ...