Related Experiment Video
Updated: Mar 8, 2026
08:47
Author Spotlight: UAV Remote Sensing for Efficient Invasive Plant Biomass Estimation
Published on: February 9, 2024
2.1K
Bag of Events: An Efficient Probability-Based Feature Extraction Method for AER Image Sensors
IEEE Transactions on Neural Networks and Learning Systems
|January 24, 2017
Summary
We introduce Bag of Events (BOE), a novel feature extraction method for Address Event Representation (AER) image sensors. BOE offers efficient, real-time processing with high accuracy, making it suitable for dynamic vision applications.
Area of Science:
- Computer Vision
- Machine Learning
- Sensor Technology
Background:
- Address Event Representation (AER) image sensors capture visual information through event sequences indicating luminance changes.
- Traditional methods may require extensive parameter tuning and pre-collected training data.
Purpose of the Study:
- To introduce a new feature extraction method for AER image sensors called Bag of Events (BOE).
- To demonstrate BOE's efficiency, accuracy, and hardware-friendliness for real-time applications.
Main Methods:
- BOE utilizes probability theory to represent objects as joint probability distributions of concurrent events.
- Each event is mapped to a unique activated pixel on the AER sensor.
- The method employs basic arithmetic operations, ensuring hardware compatibility.
Main Results:
- BOE achieves competitive real-time feature extraction speeds exceeding 275 frames/s and 120,000 events/s.
- The algorithm demonstrates robustness with a single hyperparameter, reducing tuning effort.
- Experiments on MNIST-DVS, Poker Card, and Posture datasets show BOE is faster than existing AER categorization systems with comparable accuracy.
Conclusions:
- BOE provides a mathematically interpretable, statistically robust, and computationally efficient feature extraction solution for AER sensors.
- Its online learning capability and hardware-friendliness make it ideal for real-time dynamic vision tasks.
