PulseJournal ClubResearchersJournalsExplore
Instagram
HomeJournal ClubExplore
Synapse
⌘+K
Synapse
March 4, 2026Scientific ReportsOpen Access

Hand gesture 3D pose estimation method based on swin transformer and CNN

View Full Paper
Ask AI
Bookmark
Share

Authors

RDRong DangGFGang Feng

Discussion

Loading...

Member takes

Overview

Demonstrates a novel gesture pose estimation method in depth images, suggesting enhanced prediction accuracy.

Key Points

  • The aim is to improve gesture pose estimation accuracy by addressing feature extraction and joint relationships.
  • Utilized depth images as input for gesture feature extraction.
  • Implemented a convolutional network for initial feature extraction.
  • Adopted a Swin Transformer to capture long-range joint relationships and global features.
  • Employed a U-shaped network for hierarchical feature processing and preservation of joint information.
  • Introduced 2D Gaussian heatmaps for representing keypoint distributions during feature regression.
  • Achieved an average squared error reduction from 7.012 mm to 4.776 mm compared to the baseline model.
  • Demonstrated improved performance over state-of-the-art pose estimation networks.

Cite This Study

Dang et al. (2026) studied this question.

synapsesocial.com/papers/69a7ccf7d48f933b5eed8f5ahttps://doi.org/10.1038/s41598-026-41974-6
View Full Paper
Ask AI
Bookmark
Share