Classification model construction method, classification model and object recognition method

Abstract

The embodiment of the invention relates to the technical field of deep learning, and discloses a classification model construction method, which comprises the following steps of: obtaining a feature vector of each category in N categories of known images; clustering the N categories according to the feature vectors to form a multi-layer classification hierarchical tree; taking inner nodes of the multi-layer classification hierarchical tree as nodes of the classification model, and training each node to obtain the classification model. According to the classification model construction method, the classification model and the object recognition method provided by the embodiment of the invention, the constructed classification model has a plurality of hierarchies, and the classification performance is relatively high.

Description

technical field [0001] The embodiments of the present invention relate to the technical field of deep learning, and in particular to a method for constructing a classification model, a classification model, and an object recognition method. Background technique [0002] In recent years, compared with traditional classification methods, classification methods based on deep learning have made significant breakthroughs in classification effects and high classification accuracy. As deep learning networks such as ResNet and DenseNet have been continuously proposed, based on deep learning The classification method has gradually become the main trend of classification applications. [0003] However, the inventors found at least the following problems in the prior art: when training the classification model, all categories are usually placed in the same level to train the classification model, and the resulting classification model has only one level, and the classification performa...

AI Technical Summary

Problems solved by technology

[0003] However, the inventors have found at least the following problems in the prior art: when training the classification model, all categories are usually placed on the same level to train the classification model, and the resulting classification model has only one level, and the classification performance is not high

Images