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

Recently, some researchers started exploring the use of ViTs in tackling HSIclassification and achieved remarkable results. However, the training of ViTmodels requires a considerable number of training samples, while hyperspectraldata, due to its high annotation costs, typically has a relatively small numberof training samples. This contradiction has not been effectively addressed. Inthis paper, aiming to solve this problem, we propose the single-directiontuning (SDT) strategy, which serves as a bridge, allowing us to leverageexisting labeled HSI datasets even RGB datasets to enhance the performance onnew HSI datasets with limited samples. The proposed SDT inherits the idea ofprompt tuning, aiming to reuse pre-trained models with minimal modificationsfor adaptation to new tasks. But unlike prompt tuning, SDT is custom-designedto accommodate the characteristics of HSIs. The proposed SDT utilizes aparallel architecture, an asynchronous cold-hot gradient update strategy, andunidirectional interaction. It aims to fully harness the potent representationlearning capabilities derived from training on heterologous, even cross-modaldatasets. In addition, we also introduce a novel Triplet-structured transformer(Tri-Former), where spectral attention and spatial attention modules are mergedin parallel to construct the token mixing component for reducing computationcost and a 3D convolution-based channel mixer module is integrated to enhancestability and keep structure information. Comparison experiments conducted onthree representative HSI datasets captured by different sensors demonstrate theproposed Tri-Former achieves better performance compared to severalstate-of-the-art methods. Homologous, heterologous and cross-modal tuningexperiments verified the effectiveness of the proposed SDT.

Bridging Sensor Gaps via Single-Direction Tuning for Hyperspectral Image Classification