A tiny artificial eye that spots camouflaged objects better than existing chips

An optical spiking neuron modeled on the Hodgkin-Huxley equations mimics how retinal cells encode color, adapt to brightness, and suppress background noise — all on a single device consuming as little as 0.9 picojoules per spike. By pairing it with a neural network that replicates the eye's center-surround receptive fields, the system improves detection of camouflaged objects by up to 28% over standard approaches. The result is a sensor that thinks at the edge, without shipping raw video to a server.