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

Comparing two images in a view-invariant way has been a challenging problemin computer vision for a long time, as visual features are not stable underlarge view point changes. In this paper, given a single input image of anobject, we synthesize new features for other views of the same object. Toaccomplish this, we introduce an aligned set of 3D models in the same class asthe input object image. Each 3D model is represented by a set of views, and westudy the correlation of image patches between different views, seeking what wecall surrogates --- patches in one view whose feature content predicts well thefeatures of a patch in another view. In particular, for each patch in the noveldesired view, we seek surrogates from the observed view of the given image. Fora given surrogate, we predict that surrogate using linear combination of thecorresponding patches of the 3D model views, learn the coefficients, and thentransfer these coefficients on a per patch basis to synthesize the features ofthe patch in the novel view. In this way we can create feature sets for allviews of the latent object, providing us a multi-view representation of theobject. View-invariant object comparisons are achieved simply by computing the$L^2$ distances between the features of corresponding views. We providetheoretical and empirical analysis of the feature synthesis process, andevaluate the proposed view-agnostic distance (VAD) in fine-grained imageretrieval (100 object classes) and classification tasks. Experimental resultsshow that our synthesized features do enable view-independent comparisonbetween images and perform significantly better than traditional image featuresin this respect.

3D-Assisted Image Feature Synthesis for Novel Views of an Object