Task-Oriented Semantic Feature Transmission for Multi-Task Satellite Remote Sensing over Low-SNR Channels
A task-oriented semantic transmission framework sends remote-sensing features instead of images, improving classification and detection over low-SNR satellite channels.
The paper proposes bypassing the reconstruct-then-infer paradigm in satellite remote sensing by directly transmitting semantic features extracted by a multitask-pretrained backbone. A lightweight channel adaptation module compresses feature dimensionality for bandwidth, and a feature restorer recovers task-relevant structure after channel corruption under random-SNR training. In AWGN settings, the approach consistently outperforms reconstruction-oriented JSCC baselines on scene classification and object detection, with the largest gains in the low-SNR regime.
- Transmits task-oriented semantic features instead of reconstructed images
- Channel adaptation module compresses features to reduce bandwidth
- Largest accuracy improvements occur in the low-SNR regime
From the original article
1 min read at arxiv.orgConventional satellite remote sensing transmission follows a reconstruct-then-infer paradigm that optimizes pixel-level fidelity, creating an objective mismatch with downstream tasks such as classification and detection, especially at low SNR. This paper investigates a task-oriented framework that bypasses image reconstruction and directly transmits semantic features extracted by a multitask-pretrained backbone.
By Shuoyuan Sun · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.