Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
February 12, 2026IEEE Transactions on Visualization and Computer Graphics

HiFormer: Hierarchical Transformer with Box-packed Positional Encoding for 3D Part Assembly

View Full Paper
Ask AI
Bookmark
Share

Authors

SCSongle ChenHohai UniversityLDLulu DongAnhui UniversityYZYijiao ZhouNanjing University of Posts and Telecommunications

Discussion

Loading...

Member takes

Implication

Hierarchical Transformer improves 3D part assembly accuracy in robotics and computer vision, suggesting enhanced modeling methods.

Key Points

  • The aim is to improve the estimation of 6-DoF posture in automatic 3D part assembly.
  • Developed a multi-task 3D Swin Transformer with two-stage training for better feature extraction.
  • Created a hierarchical Transformer for capturing relationships between parts at various levels.
  • Introduced box-packed positional encoding that utilizes relative box positions for improved performance.
  • Achieved average improvements of 2.84% in Part Accuracy and 3.72% in Connection Accuracy under noisy conditions.
  • Obtained 3.55% in Part Accuracy and 3.21% in Connection Accuracy under deterministic conditions.

Cite This Study

Chen et al. (2026) studied this question.

synapsesocial.com/papers/698d6d445be6419ac0d52336https://doi.org/10.1109/tvcg.2026.3662816
View Full Paper
Ask AI
Bookmark
Share

Also Consider

Synapse has enriched 3 closely related papers on similar clinical questions. Consider them for comparative context:

  1. 1Interpreting Contact Interactions to Overcome Failure in Robot Assembly Tasks2021 · 21 citations
  2. 2Factors affecting usability of 3D model learning in a virtual reality environment2019 · 52 citations
  3. 3Point Transformer2021 · 2,449 citations