Neural Network Architecture Based on Wavelet Transform for Electro-mobility Detection

Abstract

This article discusses the development of a neural network model for long-distance objects and a small-scale of computing. The proposed architecture of wavelet object detection is based on merge other frequency domains of the image into reference. The Experiments is applied for long-distance scenes to provide higher accuracy assume as lower computational complexity. Feasibility of the proposed wavelet neural networks has been evaluated and verified on vehicle detection. The architecture of wavelet object detection promoted low latency and high frame per second (fps) to porting on the NVIDIA JETSON AGX XAVIER evaluation board for artificial intelligence applications.