RetinaNet Object Detector Based on Analog-to-Spiking Neural Network Conversion

The paper proposes a method to translate a deep convolutional neural network into an equivalent spiking neural network towards the fulfillment of robust object detection in a resource-constrained platform. The aim is to provide a conversion framework that is not restricted to shallow network structures and classification problems as in state-of-the-art conversion libraries. The results show that models of higher complexity, such as the RetinaNet object detector, can be converted through rate encoding of the activations with limited loss in performance.