Image content identification method and device

Abstract

The invention provides an image content identification method and device, relates to the image identification technical field, and solves the image content identification problems in the machine vision field; the method comprises the following steps: receiving a to be indentified image; identifying the image according to an image identification model, wherein the image identification model comprises a first feature set having content information used for identifying image content; outputting an image identification result. The image content identification method and device can be applied to image content identification equipment having high accuracy requirements for image content identification.

Description

technical field [0001] Embodiments of the present invention relate to image recognition technologies, and in particular, to a method and device for recognizing image content. Background technique [0002] Image content recognition refers to interpreting the image to identify objects contained in the image, the shooting scene of the image, and the like. Image content recognition involves computer vision problems such as image-based object classification, image-based object detection, and video tracking-based relocation, and is an important part of the field of computer vision. [0003] In the prior art, a commonly used image content recognition method is an object detection method. The current object detection method mainly defines a series of templates, so that each template captures a specific type of object, or a certain part of a specific type of object, so as to capture objects that may appear in any position of the picture. Among them, the defined templates are usuall...

AI Technical Summary

Problems solved by technology

Among them, the defined templates are usually obtained by learning enough training sets, and the training sets are usually obtained by artificially marking some pictures. The collection of training sets consumes a lot of manpower and material resources, and usually cannot Obtain enough training set pictures in a short time to capture enough object category information features, and it is impossible to fully explain the complex content contained in the pictures

[0004] Therefore, the picture content recognition method of the above-mentioned prior art is inefficient, and usually cannot recognize the complex content contained in the picture

Images