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

Transformer framework has been showing superior performances in visual objecttracking for its great strength in information aggregation across the templateand search image with the well-known attention mechanism. Most recent advancesfocus on exploring attention mechanism variants for better informationaggregation. We find these schemes are equivalent to or even just a subset ofthe basic self-attention mechanism. In this paper, we prove that the vanillaself-attention structure is sufficient for information aggregation, andstructural adaption is unnecessary. The key is not the attention structure, buthow to extract the discriminative feature for tracking and enhance thecommunication between the target and search image. Based on this finding, weadopt the basic vision transformer (ViT) architecture as our main tracker andconcatenate the template and search image for feature embedding. To guide theencoder to capture the invariant feature for tracking, we attach a lightweightcorrelative masked decoder which reconstructs the original template and searchimage from the corresponding masked tokens. The correlative masked decoderserves as a plugin for the compact transform tracker and is skipped ininference. Our compact tracker uses the most simple structure which onlyconsists of a ViT backbone and a box head, and can run at 40 fps. Extensiveexperiments show the proposed compact transform tracker outperforms existingapproaches, including advanced attention variants, and demonstrates thesufficiency of self-attention in tracking tasks. Our method achievesstate-of-the-art performance on five challenging datasets, along with theVOT2020, UAV123, LaSOT, TrackingNet, and GOT-10k benchmarks. Our project isavailable at https://github.com/HUSTDML/CTTrack.

Compact Transformer Tracker with Correlative Masked Modeling