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 Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Modern change detection (CD) has achieved remarkable success by the powerfuldiscriminative ability of deep convolutions. However, high-resolution remotesensing CD remains challenging due to the complexity of objects in the scene.Objects with the same semantic concept may show distinct spectralcharacteristics at different times and spatial locations. Most recent CDpipelines using pure convolutions are still struggling to relate long-rangeconcepts in space-time. Non-local self-attention approaches show promisingperformance via modeling dense relations among pixels, yet are computationallyinefficient. Here, we propose a bitemporal image transformer (BIT) toefficiently and effectively model contexts within the spatial-temporal domain.Our intuition is that the high-level concepts of the change of interest can berepresented by a few visual words, i.e., semantic tokens. To achieve this, weexpress the bitemporal image into a few tokens, and use a transformer encoderto model contexts in the compact token-based space-time. The learnedcontext-rich tokens are then feedback to the pixel-space for refining theoriginal features via a transformer decoder. We incorporate BIT in a deepfeature differencing-based CD framework. Extensive experiments on three CDdatasets demonstrate the effectiveness and efficiency of the proposed method.Notably, our BIT-based model significantly outperforms the purely convolutionalbaseline using only 3 times lower computational costs and model parameters.Based on a naive backbone (ResNet18) without sophisticated structures (e.g.,FPN, UNet), our model surpasses several state-of-the-art CD methods, includingbetter than four recent attention-based methods in terms of efficiency andaccuracy. Our code is available at https://github.com/justchenhao/BIT\_CD.

Remote Sensing Image Change Detection With Transformers