From Event-Based Visions to Real Systems

Event: First International Workshop on Event-based Vision, ICRA 2017 · Duration: 12 min · ▶ Watch on YouTube

Abstract

Christian Brändli, CEO and co-founder of Insightness, presents the company’s journey and expertise in event-based vision technology. He highlights the extensive background of the team in developing event-based sensors and algorithms since the early days of DVS. The core message is the importance of moving beyond individual sensor components to building complete, efficient systems that integrate hardware, software, and algorithms to address real-world market needs. He demonstrates this with a collision avoidance system that outperforms a state-of-the-art drone, showcasing the practical advantages of their integrated approach.

Speakers

  • Christian Brändli — Insightness

Talks (1)

  • 00:00:00 — Christian Brändli: FROM EVENT-BASED VISIONS TO REAL SYSTEMS
    • An overview of Insightness’s history in event-based vision, emphasizing the transition from sensor technology to integrated real-world systems for commercial applications.

Key Takeaways

  • Insightness has a long-standing history and deep expertise in event-based vision, with key team members involved since the inception of DVS technology.
  • The transition from individual event-based sensors to integrated systems is crucial for commercialization and real-world applications.
  • Efficiency is the unique selling proposition (USP) of event-based technology, but it must be measured at the system level, not just the sensor level.
  • Insightness provides a complete RINO 2 Evaluation Kit that integrates sensing, processing hardware, algorithms, and software, offering various raw and processed outputs.
  • Their system demonstrates superior performance in tasks like collision avoidance compared to state-of-the-art traditional vision systems, even in challenging scenarios.

Methods / Models / Datasets Mentioned

  • DVS
  • USB 2.0 interface
  • Event-based autonomous driving
  • Color-sensitive DVS
  • Event-based structured lighting
  • DAVIS
  • BSI event sensor
  • Event lifetime estimation
  • Robust event-based stereo
  • Event-based keypoint tracking
  • EVK (Evaluation Kit)
  • DJI Phantom 4
  • RINO 2 Evaluation Kit
  • Silicon Eye sensor
  • MIPI CSI-2
  • FPGA
  • IMU
  • BlueFox Frames
  • SVO Pose
  • Visual Flow Map
  • Depth Map
  • Motion Map

Topics

Event-based vision · Dynamic Vision Sensors (DVS) · DAVIS sensors · Systems integration · Collision avoidance · Robotics · Drones · Computer vision · Hardware design · Software algorithms

Notes

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