Hardware Podcasts

How to See 'Where' Through Low-Power Event Cameras

EE Times Current · 27 Oct 2023 · 46m

What this episode covers

Professor Guillermo Gallego of the Technical University of Berlin explains how bio-inspired event-driven cameras mimic the human visual system to capture asynchronous brightness changes, offering high temporal resolution and dynamic range. He details the 'contrast maximization framework' used for motion estimation, optical flow, and depth estimation, noting that converting sparse events into frames allows compatibility with standard computer vision algorithms despite latency trade-offs. Hosts Dr. Sunny Bains and Dr. Julia D'Angelo then discuss with Professor Ralph Etienne-Cummings from Johns Hopkins University the limitations of current event cameras, particularly their lack of inter-pixel interaction compared to the biological retina, and the need for richer sensor data beyond simple motion detection.

Most relevant to

established companies

Topics

Mentioned

Companies
Technical University of Berlin, University College London, Johns Hopkins University, Italian Institute of Technology, Prophecy, DARPA, Capovacca
Products
Silicon Retina

Related episodes from other shows

Other shows on the same ground, closest first.

  • Robot Talk · 34m

    Episode Four - Robot Senses: perceiving the world

    Matt Watson of Opteran Technologies explains how reverse-engineering honeybee brains enables efficient, low-power collision avoidance and SLAM, contrasting this with the high compute demands of supervised deep learning. Dr. Perla Maiolino of the Oxford Robotics Institute discusses CySkin, a flexible, large-area artificial skin using capacitive triangular modules for tactile perception and manipulation. Dr. Dimitrios Kanoulas of UCL addresses the challenge of fusing multi-sensor data (LiDAR, cameras, IMUs) to handle uncertainty and real-time planning in dynamic, unstructured environments.

    Sensors & measurement in productsEmbedded firmware
    8 chapters

  • Robot Talk · 33m

    Episode 154: Visual navigation in insects and robots - Andrew Philippides

    Andrew Philippides, a Professor of Biorobotics at the University of Sussex, explains how studying the visual navigation of ants and bees can improve robot navigation. He details a research pipeline involving biological observation of 'learning flights' and 'learning walks', simulation, and robotics to capture natural environmental variation. The discussion highlights the 'snapshot hypothesis' and 'visual compass' algorithms, which allow robots to navigate using low-resolution images and simple view-matching rather than complex mapping. Philippides emphasizes that these bio-inspired methods are computationally efficient and robust, though challenges remain in replicating insect sensory capabilities like UV and polarized vision.

    Robotics & autonomyAI & machine learning
    8 chapters

The summary, chapters and topics are generated from the episode’s audio by this directory, not written by EE Times Current. The episode itself belongs to its publisher.