Dynamic Vision Sensors

Dynamic Vision Sensors are a type of spiking silicon retinas in which each pixel autonomously and asynchronously sends out an address event when the light it senses has changed above a given relative threshold. This type of cameras, which are "Frame-Free", do not generate sequences of still frames, as conventional commercial cameras do, but provide a flow of spiking address events that dynamically represent the changing visual scene. They are heavily inspired in biological retinas, which also send continuously nervous spike impulses to the cortex. Biological retinas are continuously vibrating through microsaccades and ocular tremors, thus producing spikes also when there is change of light. DVS cameras provide an almost instantaneous representation (with micro-second delays) of the changing visual reality, with very reduced data flow, reduced power, and data sparsity, thus reducing data processing requirements of subsequent stages. DVS cameras have become of high interest to industry recently with a number of spinoff companies commercializing them (Prophesee, IniVation, Celepixel as well as large traditional companies like Samsung and Sony embracing developments.

At IMSE there is a specific research line on AER (Address Event Representation) DVS cameras by the Neuromorphic Group, who coordinated the CAVIAR EU project in which this type of sensor was first invented and exploited. Later on they developed their own prototype which at that time had the best contrast sensitivity, power consumption, and circuit compactness, resulting in 4 licensed patents and the participation in French spinoff company Chronocam, now known as Prophesee.

Main recent activities in this line include:

  • Design and fabrication of a number of Dynamic Vision Sensors.
  • Improved AER read-out circuitry.
  • Design of improved temporal contrast sensitivity prototypes through low power mismatch-insensitive amplification stages.
  • Development of new conceptual circuits for alternative operation principles for DVS cameras.
  • Low current circuit techniques.
  • Fast read-out circuits.
Top: DVS chip with 128x128 pixels, showing zoom preview of 30µm size pixel and schematic on the right, fabricated in AMS 0.35µm. Bottom: Example captures of DVS camera showing high-speed capability, low data-rate (nev is number of events), high intra-scene dynamic range. See ref [A] for details.
3D Stereo Vision with a pair of DVS cameras solving correct object tracking with temporal occlusions. See ref [B] for details.

Contact


Bernabé Linares Barranco >
Google Scholar

Teresa Serrano Gotarredona >
Google Scholar

Luis A. Camuñas Mesa >
Google Scholar

Group of Neuromorphic Systems

Keywords


  • Dynamic Vision Sensor
  • Address Event Representation
  • Spiking Retinas
  • Spiking Neural Networks
  • Asynchronous Circuits
  • High-Speed Low-Power Vision
  • DVS Stereo-Vision

Research Highlights


A. Yousefzadeh, G. Orchard, T. Serrano-Gotarredona and B. Linares-Barranco, "Active Perception with Dynamic Vision Sensors. Minimum Saccades with Optimum Recognition", IEEE Transactions on Biomedical Circuits and Systems, vol. 12, no. 4, pp 927-939, 2018  »  doi

[B] L.A. Camuñas-Mesa, T. Serrano-Gotarredona, S. Ieng, R. Benosman and B. Linares-Barranco, "Event-Driven Stereo Visual Tracking Algorithm to Solve Object Occlusion", IEEE Transactions on Neural Networks and Learning Systems, vol. 29, no. 9, pp 4223-4237, 2017  »  doi

T. Serrano-Gotarredona and B. Linares-Barranco, "Poker-DVS and MNIST-DVS. Their History, How They were Made, and Other Details", Frontiers in Neuromorphic Engineering, Frontiers in Neuroscience, vol. 9, article 481, 2015  »  doi

[A] T. Serrano-Gotarredona and B. Linares-Barranco, "A 128x128 1.5% Contrast Sensitivity 0.9% FPN 3µs Latency 4mW Asynchronous Frame-Free Dynamic Vision Sensor Using Transimpedance Amplifiers", IEEE Journal of Solid-State Circuits, vol. 48, no. 3, pp 827-838, 2013  »  doi

J.A. Leñero-Bardallo, T. Serrano-Gotarredona and B. Linares-Barranco, "A 3.6µs Latency Asynchronous Frame-Free Event-Driven Dynamic-Vision-Sensor", IEEE Journal of Solid-State Circuits, vol. 46, no. 6, pp 1443-1455, 2011  »  doi

Technology Transfer


Patent. T. Finateu, B. Linares-Barranco, C. Posch and T. Serrano-Gotarredona, "Pixel Circuit for Detecting Time-Dependent Visual Data", WO2018073379A1. Priority 20-Oct-2016. European patent, extended to US, Korea, Japan, China.

Patent. T. Finateu, B. Linares-Barranco, C. Posch and T. Serrano-Gotarredona, "Sample and Hold based Temporal Contrast Vision Sensor", WO2017174579A1. Priority: 4-Apr-2016.

Patent. B. Linares-Barranco and T. Serrano-Gotarredona, "Method and Device for Detecting the Temporal Variation of the Light Intensity in a Matrix of Photosensors", WO2014091040A1. Priority: 11-Dec-2012. European patent, extended to US, Korea, Japan, Israel.

Patent. B. Linares-Barranco and T. Serrano-Gotarredona, "Low-Mismatch and Low-Consumption Transimpedance Gain Circuit for Temporally Differentiating Phot-Sensing systems in dynamic vision Sensors", WO2012160230A1. Priority: 26-May-2011. European patent, extended to US, Korea, Japan, China.

Prophesee
Spin-off Company. Metavision for machines

Key Research Projects & Contracts


APROVIS3D: Analog PROcessing Of Bioinspired Vision Sensors For 3D Reconstruction
PI: Teresa Serrano Gotarredona
Funding Body: Min. de Ciencia e Innovación
Apr 2020 - March 2023

COGNET: Event-based cognitive vision system. Extension to audio with sensory fusion
PI: Teresa Serrano Gotarredona
Funding Body: Min. de Ciencia e Innovación
Jan 2016 - Dec 2019

ECOMODE: Event-driven compressive vision for multimodal interaction with mobile devices
PI: Bernabé Linares-Barranco
Funding Body: European Union
Jan 2015 - Dec 2018
WEBSITE

BIOSENSE: Bioinspired event-based system for sensory fusion and neurocortical processing. High-speed low-cost applications in robotics and automotion
PI: Teresa Serrano Gotarredona
Funding Body: Min. de Ciencia e Innovación
Jan 2013 - Dec 2015

NANONEURO: Design of neurocortical architectures for vision applications
PI: Teresa Serrano Gotarredona
Funding Body: Junta de Andalucía
Jul 2011 - Dec 2014

All Research Areas & Lines


SHARE