Related Experiment Video
Updated: Aug 9, 2026
10:56
Long-term Behavioral Tracking of Freely Swimming Weakly Electric Fish
Published on: March 6, 2014
12.5K
SiamEFT: adaptive-time feature extraction hybrid network for RGBE multi-domain object tracking
Shuqi Liu1, Gang Wang2, Yong Song1
1School of Optics and Photonics, Beijing Institute of Technology, Beijing, China.
Frontiers in Neuroscience
|August 23, 2024
Summary
We developed a novel Siamese Event Frame Tracker (SiamEFT) for robust object tracking using RGB and Event (RGBE) data. SiamEFT enhances spatio-temporal feature utilization, significantly improving accuracy and efficiency in complex scenes.
Area of Science:
- Computer Vision
- Robotics
- Artificial Intelligence
Background:
- Object tracking often fails with complex backgrounds.
- Existing RGBE tracking methods neglect unique spatio-temporal features.
- Robust tracking requires effective integration of diverse data domains.
Purpose of the Study:
- To propose a novel tracker, SiamEFT, for effective RGBE feature utilization.
- To address limitations in current RGBE object tracking methods.
- To enhance tracking accuracy and efficiency in challenging scenarios.
Main Methods:
- Developed an adaptive-time attention module for event data aggregation.
- Designed a hybrid network combining artificial and spiking neural networks.
- Implemented cross-network fusion for comprehensive spatio-temporal feature extraction.
Main Results:
- SiamEFT achieved success rates of 0.456 and 0.574 on VisEvent and COESOT datasets.
- Outperformed state-of-the-art competing methods in object tracking.
- Demonstrated a 2.3-fold enhancement in tracking efficiency.
Conclusions:
- SiamEFT offers superior accuracy and efficiency for RGBE object tracking.
- The proposed method effectively utilizes diverse spatio-temporal features.
- Validated effectiveness in challenging and diverse visual scenes.
