Method, apparatus and computing device for object recognition neural network

摘要

To provide a training method of an object recognition neural network, an apparatus, and a computing device, which makes it possible to recognize a distorted fisheye image.SOLUTION: A training method of an object recognition neural network includes the steps of: acquiring a distorted fisheye image including at least one target object; defining the distorted fisheye image in several regions according to a degree of distortion; determining a region to which each target object belongs; classifying each target object according to the region to which each target object belongs; and inputting the distorted fisheye image and a category to which each target object belongs into the object recognition neural network to train the object recognition neural network. The training method can enhance a recognition effect of target object recognition by the object recognition neural network on the basis of the distorted fisheye image.SELECTED DRAWING: Figure 1