English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

In recent years, remote sensing (RS) vision foundation models such as RingMohave emerged and achieved excellent performance in various downstream tasks.However, the high demand for computing resources limits the application ofthese models on edge devices. It is necessary to design a more lightweightfoundation model to support on-orbit RS image interpretation. Existing methodsface challenges in achieving lightweight solutions while retaininggeneralization in RS image interpretation. This is due to the complex high andlow-frequency spectral components in RS images, which make traditional singleCNN or Vision Transformer methods unsuitable for the task. Therefore, thispaper proposes RingMo-lite, an RS multi-task lightweight network with aCNN-Transformer hybrid framework, which effectively exploits thefrequency-domain properties of RS to optimize the interpretation process. It iscombined by the Transformer module as a low-pass filter to extract globalfeatures of RS images through a dual-branch structure, and the CNN module as astacked high-pass filter to extract fine-grained details effectively.Furthermore, in the pretraining stage, the designed frequency-domain maskedimage modeling (FD-MIM) combines each image patch's high-frequency andlow-frequency characteristics, effectively capturing the latent featurerepresentation in RS data. As shown in Fig. 1, compared with RingMo, theproposed RingMo-lite reduces the parameters over 60% in various RS imageinterpretation tasks, the average accuracy drops by less than 2% in most of thescenes and achieves SOTA performance compared to models of the similar size. Inaddition, our work will be integrated into the MindSpore computing platform inthe near future.

RingMo-lite: A Remote Sensing Multi-task Lightweight Network with CNN-Transformer Hybrid Framework