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

High-resolution representations (HR) are essential for dense prediction taskssuch as segmentation, detection, and pose estimation. Learning HRrepresentations is typically ignored in previous Neural Architecture Search(NAS) methods that focus on image classification. This work proposes a novelNAS method, called HR-NAS, which is able to find efficient and accuratenetworks for different tasks, by effectively encoding multiscale contextualinformation while maintaining high-resolution representations. In HR-NAS, werenovate the NAS search space as well as its searching strategy. To betterencode multiscale image contexts in the search space of HR-NAS, we firstcarefully design a lightweight transformer, whose computational complexity canbe dynamically changed with respect to different objective functions andcomputation budgets. To maintain high-resolution representations of the learnednetworks, HR-NAS adopts a multi-branch architecture that provides convolutionalencoding of multiple feature resolutions, inspired by HRNet. Last, we proposedan efficient fine-grained search strategy to train HR-NAS, which effectivelyexplores the search space, and finds optimal architectures given various tasksand computation resources. HR-NAS is capable of achieving state-of-the-arttrade-offs between performance and FLOPs for three dense prediction tasks andan image classification task, given only small computational budgets. Forexample, HR-NAS surpasses SqueezeNAS that is specially designed for semanticsegmentation while improving efficiency by 45.9%. Code is available athttps://github.com/dingmyu/HR-NAS

HR-NAS: Searching Efficient High-Resolution Neural Architectures with Lightweight Transformers