Fig. 4: Comprehensive evaluation of MSF neurons on large-scale object detection tasks.
From: A multisynaptic spiking neuron for simultaneously encoding spatiotemporal dynamics
a Performance-power comparison of different models and neurons on the COCO dataset. b Detection visualization comparison of LIF and MSF neurons on the COCO dataset. c Event stream data of one sample from Gen1 and the corresponding accumulated frame images. d Performance-power comparison of different models and neurons on Gen1. e Detection visualization comparison of LIF and MSF neurons on the Gen1 dataset. To elucidate the influence of timesteps on detection capability, we have varied the timesteps of MSF neurons to 1, 3, and 6, respectively.
