Method, apparatus and computing device for object recognition neural network

摘要

Problem to be solved: to provide an object recognition neural network training method, apparatus and computing device capable of recognizing distorted fish eye images.How to train object recognition neural networksObtaining a distorted fish eye image containing at least one target object andThe process of forming a distorted fish eye image in several regions according to the degree of distortionDetermining the area in which each target object belongsClassify each target object according to the area in which each target object belongsA distorted fish eye image and a category belonging to each target object are input to an object recognition neural networkTraining object recognition neural networks and.The recognition effect of target object recognition can be enhanced by object recognition neural network based on distorted fish eye image.Diagram