Event Computer Vision 10 years Assessment: Where We Came From, Where We Are and Where We Are Heading To

Event: CVPR 2025 Workshop on Event-based Vision · Duration: 27 min · ▶ Watch on YouTube

Abstract

This talk provides a comprehensive overview of event-based computer vision, tracing its origins from early neuromorphic engineering efforts inspired by biological vision systems, particularly the retina. It highlights the current state where event cameras have become a commodity, with major tech companies and startups actively developing and integrating them. The presentation then delves into the inherent advantages of event-based sensing, such as sparse data representation and high temporal precision, contrasting it with the inefficiencies of traditional frame-based systems. Finally, it outlines future directions, advocating for dedicated processors, exploring new acquisition methods, and fostering a deeper understanding of neurosciences to unlock the full potential of event cameras for low-power, low-latency applications in areas like IoT and advanced robotics.

Speakers

  • R.B. Benosman — University of Pittsburgh, Carnegie Mellon, Sorbonne Universitas, Eye & Ear and McGowan Institute

Talks (1)

  • 00:00:00 — R.B. Benosman: Event Computer Vision 10 years Assessment: Where We Came From, Where We Are and Where We Are Heading To
    • An assessment of the past, present, and future of event-based computer vision, emphasizing its origins in neuromorphic engineering, current commodity status, and future potential for low-power, low-latency applications by moving beyond frame-based thinking and improving hardware architecture.

Key Takeaways

  • Event-based vision originated from neuromorphic efforts to mimic biological systems like the retina, offering inherent advantages in sparse data and temporal precision.
  • Event cameras have transitioned from niche research tools to a commercial commodity, with widespread adoption by major tech companies and startups.
  • Traditional frame-based processing is inefficient for event data, leading to high power consumption and latency due to over-sampling and the Von Neumann bottleneck.
  • The future of event-based vision lies in developing dedicated processors and exploring new acquisition methods that fully leverage the temporal properties of events, moving beyond direct biological replication to find optimal engineering abstractions.
  • A new generation of engineers with a strong understanding of neurosciences is crucial to advance event-based vision for low-power, low-latency applications like IoT.

Methods / Models / Datasets Mentioned

  • DVS
  • ATIS
  • Faster RCNN
  • Viola Jones
  • SSD
  • Von Neumann architecture
  • Qualcomm Zeroth
  • IBM TrueNorth
  • Intel Loihi
  • BrainChip

Topics

Event-based vision · Neuromorphic engineering · Retina · Event cameras · Low power · Low latency · Incremental computation · Hardware architecture · Biological inspiration · IoT

Notes

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