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

We study the composition style in deep image matting, a notion thatcharacterizes a data generation flow on how to exploit limited foregrounds andrandom backgrounds to form a training dataset. Prior art executes this flow ina completely random manner by simply going through the foreground pool or byoptionally combining two foregrounds before foreground-background composition.In this work, we first show that naive foreground combination can beproblematic and therefore derive an alternative formulation to reasonablycombine foregrounds. Our second contribution is an observation that mattingperformance can benefit from a certain occurrence frequency of combinedforegrounds and their associated source foregrounds during training. Inspiredby this, we introduce a novel composition style that binds the source andcombined foregrounds in a definite triplet. In addition, we also find thatdifferent orders of foreground combination lead to different foregroundpatterns, which further inspires a quadruplet-based composition style. Resultsunder controlled experiments on four matting baselines show that ourcomposition styles outperform existing ones and invite consistent performanceimprovement on both composited and real-world datasets. Code is available at:https://github.com/coconuthust/composition_styles

Infusing Definiteness into Randomness: Rethinking Composition Styles for Deep Image Matting