Model training method and device, image category detection method and device and electronic equipment

Abstract

The embodiment of the invention discloses a model training method and device, an image category detection method and device and electronic equipment. An embodiment of the method comprises the steps ofobtaining a sample set; determining a target sampling probability of each category of image samples based on the number of each category of image samples, and enabling the number of each category ofimage samples to be inversely proportional to the target sampling probability; using a random sampling mode and a sampling mode based on a target sampling probability respectively for sampling image samples in the sample set, taking the image samples sampled in different sampling modes as inputs of different branch networks in the initial model containing the double-branch network, and training the initial model based on annotation information of the sampled image samples to obtain an image category detection model. According to the embodiment, the data utilization rate in the model training process is improved, so that the model fully learns the characteristics of various types of image samples, and the detection precision of the image type detection model is improved.

Description

technical field [0001] The embodiments of the present application relate to the field of computer technology, and specifically relate to a model training method, an image category detection method, an apparatus, and electronic equipment. Background technique [0002] Image recognition is a technology that uses computers to process, analyze and understand images to identify targets and objects in various patterns. When using image recognition technology to detect the category of an image, a machine learning method can usually be used to train an image category detection model, so that the model can be used to detect the category of the image to be tested. [0003] In the prior art, in the process of training the image category detection model, random sampling is usually used to sample the image samples in the sample set, so that each image sample has the same sampling probability, and then the sequentially sampled image samples are used for image category detection Model tra...

AI Technical Summary

Problems solved by technology

However, the image samples in the sample set usually have a long-tail problem, that is, the number of image samples in a small number of categories is large, and the number of image samples in most categories is small, resulting in an unbalanced number of image samples in different categories in the sample set.

In this case, for categories with a small number of image samples, sufficient image samples cannot be sampled for these categories, resulting in insufficient feature learning of image samples of these categories, which in turn results in low detection accuracy of the image category detection model

Images