Related Experiment Video
Updated: May 24, 2025
06:37
Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention
Published on: December 15, 2023
2.5K
Object Recognition Using Shape and Texture Tactile Information: A Fusion Network Based on Data Augmentation and
IEEE Transactions on Haptics
|March 3, 2025
Summary
This study introduces TSMFormer, an attention-based network for tactile object recognition. It effectively integrates shape and texture, significantly improving recognition accuracy for complex objects.
Area of Science:
- Robotics
- Computer Vision
- Artificial Intelligence
Background:
- Current tactile object recognition often relies on single attributes (shape or texture), leading to poor performance with similar objects.
- Limited research exists on integrating shape and texture, and existing fusion methods overlook feature interactions.
Purpose of the Study:
- To develop a novel attention-based fusion network, TSMFormer, for enhanced tactile object recognition.
- To explore feature interactions between shape and texture using attention mechanisms in tactile images.
Main Methods:
- Proposed TSMFormer, an attention-based fusion network integrating shape and texture information.
- Expanded the tactile image dataset using data augmentation to leverage Transformer network capabilities.
- Conducted comparative experiments to evaluate the network's performance against existing methods.
Main Results:
- Achieved a significant accuracy improvement to 99.3% by combining texture and shape information.
- Demonstrated the effectiveness of the proposed attention fusion mechanism compared to existing fusion methods.
- Validated TSMFormer's capability in fusing texture and shape information via an attention mechanism.
Conclusions:
- TSMFormer offers a valuable approach for fusing texture and shape information in tactile images using attention.
- The network shows significant potential for practical applications like robot grasping and industrial quality inspection.
