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

Numerous advancements in deep learning can be attributed to the access tolarge-scale and well-annotated datasets. However, such a dataset isprohibitively expensive in 3D computer vision due to the substantial collectioncost. To alleviate this issue, we propose a cost-effective method forautomatically generating a large amount of 3D objects with annotations. Inparticular, we synthesize objects simply by assembling multiple randomprimitives. These objects are thus auto-annotated with part labels originatingfrom primitives. This allows us to perform multi-task learning by combining thesupervised segmentation with unsupervised reconstruction. Considering the largeoverhead of learning on the generated dataset, we further propose a datasetdistillation strategy to remove redundant samples regarding a target dataset.We conduct extensive experiments for the downstream tasks of 3D objectclassification. The results indicate that our dataset, together with multi-taskpretraining on its annotations, achieves the best performance compared to othercommonly used datasets. Further study suggests that our strategy can improvethe model performance by pretraining and fine-tuning scheme, especially for thedataset with a small scale. In addition, pretraining with the proposed datasetdistillation method can save 86\% of the pretraining time with negligibleperformance degradation. We expect that our attempt provides a new data-centricperspective for training 3D deep models.

Primitive3D: 3D Object Dataset Synthesis from Randomly Assembled Primitives