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 Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Estimating 6D poses and reconstructing 3D shapes of objects in open-worldscenes from RGB-depth image pairs is challenging. Many existing methods rely onlearning geometric features that correspond to specific templates whiledisregarding shape variations and pose differences among objects in the samecategory. As a result, these methods underperform when handling unseen objectinstances in complex environments. In contrast, other approaches aim to achievecategory-level estimation and reconstruction by leveraging normalized geometricstructure priors, but the static prior-based reconstruction struggles withsubstantial intra-class variations. To solve these problems, we propose theDTF-Net, a novel framework for pose estimation and shape reconstruction basedon implicit neural fields of object categories. In DTF-Net, we design adeformable template field to represent the general category-wise shape latentfeatures and intra-category geometric deformation features. The fieldestablishes continuous shape correspondences, deforming the category templateinto arbitrary observed instances to accomplish shape reconstruction. Weintroduce a pose regression module that shares the deformation features andtemplate codes from the fields to estimate the accurate 6D pose of each objectin the scene. We integrate a multi-modal representation extraction module toextract object features and semantic masks, enabling end-to-end inference.Moreover, during training, we implement a shape-invariant training strategy anda viewpoint sampling method to further enhance the model's capability toextract object pose features. Extensive experiments on the REAL275 and CAMERA25datasets demonstrate the superiority of DTF-Net in both synthetic and realscenes. Furthermore, we show that DTF-Net effectively supports grasping taskswith a real robot arm.

DTF-Net: Category-Level Pose Estimation and Shape Reconstruction via Deformable Template Field