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

Existing CNN-based methods for pixel labeling heavily depend on multi-scalefeatures to meet the requirements of both semantic comprehension and detailpreservation. State-of-the-art pixel labeling neural networks widely exploitconventional scale-transfer operations, i.e., up-sampling and down-sampling tolearn multi-scale features. In this work, we find that these operations lead toscale-confused features and suboptimal performance because they arespatial-invariant and directly transit all feature information cross scaleswithout spatial selection. To address this issue, we propose the GatedScale-Transfer Operation (GSTO) to properly transit spatial-filtered featuresto another scale. Specifically, GSTO can work either with or without extrasupervision. Unsupervised GSTO is learned from the feature itself while thesupervised one is guided by the supervised probability matrix. Both forms ofGSTO are lightweight and plug-and-play, which can be flexibly integrated intonetworks or modules for learning better multi-scale features. In particular, byplugging GSTO into HRNet, we get a more powerful backbone (namely GSTO-HRNet)for pixel labeling, and it achieves new state-of-the-art results on the COCObenchmark for human pose estimation and other benchmarks for semanticsegmentation including Cityscapes, LIP and Pascal Context, with negligibleextra computational cost. Moreover, experiment results demonstrate that GSTOcan also significantly boost the performance of multi-scale feature aggregationmodules like PPM and ASPP. Code will be made available athttps://github.com/VDIGPKU/GSTO.

GSTO: Gated Scale-Transfer Operation for Multi-Scale Feature Learning in Pixel Labeling