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

In real-world applications, it is essential to jointly estimate the 3D objectpose and class label of objects, i.e., to perform 3D-aware classification.Whilecurrent approaches for either image classification or pose estimation can beextended to 3D-aware classification, we observe that they are inherentlylimited: 1) Their performance is much lower compared to the respectivesingle-task models, and 2) they are not robust in out-of-distribution (OOD)scenarios. Our main contribution is a novel architecture for 3D-awareclassification, which builds upon a recent work and performs comparably tosingle-task models while being highly robust. In our method, an object categoryis represented as a 3D cuboid mesh composed of feature vectors at each meshvertex. Using differentiable rendering, we estimate the 3D object pose byminimizing the reconstruction error between the mesh and the featurerepresentation of the target image. Object classification is then performed bycomparing the reconstruction losses across object categories. Notably, theneural texture of the mesh is trained in a discriminative manner to enhance theclassification performance while also avoiding local optima in thereconstruction loss. Furthermore, we show how our method and feed-forwardneural networks can be combined to scale the render-and-compare approach tolarger numbers of categories. Our experiments on PASCAL3D+, occluded-PASCAL3D+,and OOD-CV show that our method outperforms all baselines at 3D-awareclassification by a wide margin in terms of performance and robustness.

Robust 3D-aware Object Classification via Discriminative Render-and-Compare