Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
May 24, 2026ActuatorsOpen Access

ADDFNet: A Robotic Grasping Depth Map Completion Network Integrating Differential Enhancement Convolution and Hybrid Attention

View Full Paper
Ask AI
Bookmark
Share

Authors

NLNan LiuYLYi-Horng LaiYWYue Wu

Discussion

Loading...

Member takes

Overview

Randomized trial shows improved accuracy in depth completion for transparent objects, indicating better robotic grasping performance.

Key Points

  • The central aim is to improve depth completion for transparent objects to aid in robotic grasping tasks.
  • Developed ADDFNet, featuring Multi-directional Differential Attention Module (MDAM) and Cross-Modal Feature Refinement (CMFR).
  • Implemented detail enhancement through multi-branch differential convolution and dynamic convolution with symmetry-enhanced geometry attention.
  • Tested against ClearPose and TransCG datasets for performance evaluation.
  • ADDFNet outperformed existing methods in RMSE, REL, MAE, and threshold accuracy metrics.
  • Demonstrated superior stability in edge recovery and detail reconstruction of transparent objects.
  • Achieved improved accuracy and robustness through advanced feature extraction and interaction techniques.

Cite This Study

Liu et al. (2026) studied this question.

synapsesocial.com/papers/6a12969048a0ea16656737d8https://doi.org/10.3390/act15060280
View Full Paper
Ask AI
Bookmark
Share

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

  1. 1Transparent Object Depth Completion2024
  2. 2Transparent Depth Completion Using Segmentation Features2024
  3. 3SRNet-Trans: A Singal-Image Guided Depth Completion Regression Network for Transparent Object2025 · 2 citations
  4. 4SRNet-Trans：A Signal-Image Guided Depth Completion Regression Network for Transparent Object2025 · 1 citations
  5. 5DistillGrasp: Integrating Features Correlation with Knowledge Distillation for Depth Completion of Transparent Objects2024