Neuromorphic Decoding and Dynamical Image Reconstruction from Populations of Spiking Retinal Ganglion Cells

Key points

• Natural vision is dynamic, yet state-of-the-art decoders mostly reconstruct single frames with artificial neural networks that process dense, continuous-valued signals rather than the retina's sparse, asynchronous spikes.1
• We asked whether dynamic scenes can be reconstructed and tracked entirely within the spike domain, and at what computational cost.1
• We built an end-to-end spiking neural network (SNN) whose neurons integrate the recorded RGC spikes over time.1
• Populations of up to 1836 RGCs were recorded ex vivo from mouse retina on high-density microelectrode arrays while flashed images, moving bars, drifting gratings and randomly moving handwritten digits were presented.1

Sources (1)

• [1]Neuromorphic Decoding and Dynamical Image Reconstruction from Populations of Spiking Retinal Ganglion Cells
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 12, 05:33 PM

  Decoding visual stimuli from the spiking activity of retinal ganglion cells (RGCs) informs both our understanding of the retinal code and the design of visual prostheses.

  Natural vision is dynamic, yet state-of-the-art decoders mostly reconstruct single frames with artificial neural networks that process dense, continuous-valued signals rather than the retina's sparse, asynchronous spikes.

Extractive summary: sentences quoted from the sources.