ABSTRACT
This chapter presents methodologies and optimization techniques for deploying SNNs on neuromorphic hardware. The Intel Loihi processor is used as a target computing platform to deploy energy-efficient SNNs. An overview of the Loihi chip is discussed in Section 6.1. Afterward, different applications were treated and deployed. Section 6.2 presents a methodology for implementing efficient SNNs for gesture recognition. First, the analysis to optimize the DNN-to-SNN conversion is discussed in Section 6.2.2. Then, the pre-processing method for enabling the training of event data in the DNN domain is presented in Section 6.2.3. Towards applications more oriented to autonomous cars, Section 6.3 presents efficient implementations of event-based SNNs for car recognition, while Section 6.4 presents SNNs for lane detection implemented on the Loihi neuromorphic processor.
