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 propose a simple and efficient method for exploiting synthetic images whentraining a Deep Network to predict a 3D pose from an image. The ability ofusing synthetic images for training a Deep Network is extremely valuable as itis easy to create a virtually infinite training set made of such images, whilecapturing and annotating real images can be very cumbersome. However, syntheticimages do not resemble real images exactly, and using them for training canresult in suboptimal performance. It was recently shown that for exemplar-basedapproaches, it is possible to learn a mapping from the exemplar representationsof real images to the exemplar representations of synthetic images. In thispaper, we show that this approach is more general, and that a network can alsobe applied after the mapping to infer a 3D pose: At run time, given a realimage of the target object, we first compute the features for the image, mapthem to the feature space of synthetic images, and finally use the resultingfeatures as input to another network which predicts the 3D pose. Since thisnetwork can be trained very effectively by using synthetic images, it performsvery well in practice, and inference is faster and more accurate than with anexemplar-based approach. We demonstrate our approach on the LINEMOD dataset for3D object pose estimation from color images, and the NYU dataset for 3D handpose estimation from depth maps. We show that it allows us to outperform thestate-of-the-art on both datasets.

Feature Mapping for Learning Fast and Accurate 3D Pose Inference from Synthetic Images