A Real-time Detection Embedded Edge Computing System Based on SNN-ONNX Model Deployment Inference

With the development of neural network technology, Spiking Neural Networks (SNNs) have shown great potential in edge computing and embedded systems due to their biologically inspired and low-power characteristics. This paper proposes an optimization method based on a custom LIF neuron model to solve the problem of operator conflicts that arise during the conversion of trained SNNs models to ONNX format. By replacing the official LIF neuron model with a custom LIF neuron and achieving cross-framework deployment after model conversion, the method was successfully applied to an embedded system for real-time image classification tasks, meeting the real-time detection requirements in edge computing environments. This method not only completes the hardware implementation of the SNNs model, but also significantly improves its portability and interoperability in embedded systems.