Event-Based SLAM at Slamcore

Event: CVPR 2017 Workshop on Event-Based Vision · Duration: 14 min · ▶ Watch on YouTube

Abstract

Slamcore, a spinout from Imperial College London, is developing modular SLAM systems that integrate traditional and novel sensors, processors, and algorithms. This presentation introduces their work on event-based SLAM, highlighting the potential of event cameras for handling rapid motion, high dynamic range, low latency, and low power consumption. They demonstrate real-time event-based 3-DoF tracking and 2D mapping (ETAM 2D) and 6-DoF tracking and 3D mapping (ETAM 3D), and discuss the challenges and opportunities in processing high event rates from advanced event cameras for practical applications.

Speakers

  • Hannes K. — Slamcore

Talks (1)

  • 00:00:00 — Hannes K.: Event-Based SLAM at Slamcore
    • An overview of Slamcore’s approach to SLAM, focusing on the integration of traditional and novel sensors like event cameras, and addressing challenges in rapid motion, high dynamic range, low latency, and low power consumption.

Key Takeaways

  • Slamcore aims to develop modular SLAM systems by innovating across sensors, processors, and algorithms.
  • Event cameras offer significant advantages for SLAM, particularly in scenarios with rapid motion, high dynamic range, and strict power/latency requirements.
  • ETAM 2D and ETAM 3D are real-time event-based SLAM methods developed by Slamcore for 3-DoF and 6-DoF tracking and mapping.
  • A key challenge for event-based SLAM is efficiently processing the high event rates generated by newer, higher-resolution event cameras on computationally limited platforms.
  • Combining event cameras with IMUs can enhance robustness and accuracy for pose estimation.

Methods / Models / Datasets Mentioned

  • BRISK
  • OKVIS
  • Multi-scale Monocular Fusion
  • MonoSLAM
  • PTAM
  • ORB-SLAM
  • ORB-SLAM++
  • ETAM 2D
  • ETAM 3D
  • Intel Movidius 2
  • iniLabs DVS128
  • Insightness SER1
  • Samsung DVS Gen.2

Topics

Event-Based SLAM · Simultaneous Localisation And Mapping (SLAM) · Event cameras · Visual SLAM · Sensor fusion · Real-time tracking · 3D mapping · Low power platforms · High dynamic range · Rapid motion

Notes

Open for commentary — connections to other work, critiques, follow-up reading.