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/

Zendy Plus

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

Zendy Tools

Zendy Open

The impressive performance of deep convolutional neural networks insingle-view 3D reconstruction suggests that these models perform non-trivialreasoning about the 3D structure of the output space. However, recent work haschallenged this belief, showing that complex encoder-decoder architecturesperform similarly to nearest-neighbor baselines or simple linear decoder modelsthat exploit large amounts of per category data in standard benchmarks. On theother hand settings where 3D shape must be inferred for new categories with fewexamples are more natural and require models that generalize about shapes. Inthis work we demonstrate experimentally that naive baselines do not apply whenthe goal is to learn to reconstruct novel objects using very few examples, andthat in a \emph{few-shot} learning setting, the network must learn conceptsthat can be applied to new categories, avoiding rote memorization. To addressdeficiencies in existing approaches to this problem, we propose threeapproaches that efficiently integrate a class prior into a 3D reconstructionmodel, allowing to account for intra-class variability and imposing an implicitcompositional structure that the model should learn. Experiments on the popularShapeNet database demonstrate that our method significantly outperform existingbaselines on this task in the few-shot setting.

Few-Shot Single-View 3-D Object Reconstruction with Compositional Priors