Neuromorphic Decoding and Dynamical Image Reconstruction from Populations of Spiking Retinal Ganglion Cells
Researchers built a spiking neural network (SNN) to decode dynamic visual scenes directly from the spiking activity of retinal ganglion cells (RGCs). The SNN successfully reconstructed flashed images with high similarity and tracked moving objects accurately. Crucially, this spike-based method used significantly less computational energy than traditional artificial networks, suggesting viability for low-power implants.