Image recognition model training method and device and electronic equipment

Abstract

The invention provides an image recognition model training method and device and electronic equipment, and relates to the technical field of deep learning, and the method comprises the steps: inputting a training sample marked with a sample label in advance into an image recognition model; in an iterative training process of an image recognition model, determining a fine-grained feature map corresponding to the training sample based on a network layer of the image recognition model, inputting the fine-grained feature map into a preset deep learning network to enable the deep learning network to learn fine-grained feature information from the fine-grained feature map, and distilling the learned fine-grained feature information into the image recognition model; wherein the fine-grained feature map is an image marked with a discriminative region corresponding to the sample label; and repeatedly executing the above training steps until the training is finished, and obtaining a trained image recognition model. According to the invention, the image recognition efficiency of the trained image recognition model can be improved.

Description

technical field [0001] The present invention relates to the technical field of deep learning, in particular to a training method, device and electronic equipment for an image recognition model. Background technique [0002] In the existing human action recognition tasks, when training the image recognition model, in order to obtain the detailed features of the training samples, it is necessary to accurately locate the discriminative area (ie, the area where the action occurs) in the sample image. Traditional image recognition The model training method mainly relies on manual labeling of discriminative regions, which is time-consuming and laborious. In order to save the time of manual labeling of discriminative regions, researchers began to use self-supervised attention mechanism to mine discriminative regions to locate discriminative regions in sample images, thus To obtain fine-grained features, however, when using the attention mechanism to mine discriminative regions for ...

AI Technical Summary

Problems solved by technology

[0002] In the existing human action recognition tasks, when training the image recognition model, in order to obtain the detailed features of the training samples, it is necessary to accurately locate the discriminative area (ie, the area where the action occurs) in the sample image. Traditional image recognition The model training method mainly relies on manual labeling of discriminative regions, which is time-consuming and laborious. In order to save the time of manual labeling of discriminative regions, researchers began to use self-supervised attention mechanism to mine discriminative regions to locate discriminative regions in sample images, thus To obtain fine-grained features, however, when using the attention mechanism to mine discriminative regions for model training, it is necessary to use multiple models, and when using the trained model for image recognition, it is also necessary to use multiple models for staged recognition, and the amount of calculation is relatively large. Big

Therefore, the image recognition model obtained by the existing model training method still has the problem of low image recognition efficiency due to the large amount of image recognition calculations.

Images