Image recognition method and device and image recognition model training method and device

Abstract

The invention discloses an image recognition method. The method comprises the steps of obtaining a to-be-recognized image; obtaining a first image feature of the to-be-identified image through a small-scale image identification model, the small-scale image identification model being deployed in a terminal device; determining image similarity between the first image feature and the second image feature according to the first image feature and the N second image features, the second image feature being an image feature acquired by the to-be-matched image through a large-scale image recognition model; and determining an image recognition result of the to-be-recognized image according to the image similarity. The invention further discloses an image recognition model training method and device. High-quality image features are extracted through the large-scale image recognition model, efficient calculation can be carried out through the small-scale image recognition model, and therefore therecognition accuracy of the small-scale image recognition model is improved on the premise that the operation efficiency is guaranteed.

Description

technical field [0001] The present application relates to the field of artificial intelligence, in particular to an image recognition method, a method and a device for image recognition model training. Background technique [0002] Face recognition is an important research topic in the field of computer vision and is widely used in industry. With the development and popularization of mobile devices, the demand for running face recognition algorithms on terminal devices is increasing. However, the limited computing power, storage space, and high requirements for real-time computing of the terminal system make it impossible to run large-scale neural network models directly on it. [0003] At present, the face recognition method designed for terminal equipment is to improve the structure and operation of the large-scale face recognition convolutional neural network (Convolutional Neural Networks, CNN) model, while maintaining the performance of the model as much as possible. ...

AI Technical Summary

Problems solved by technology

However, due to the limited model parameters of the small model, the complexity of the solution that can be fitted is much smaller than that of the large model, resulting in a decrease in recognition accuracy

If the large model is directly applied to the terminal system, although high recognition performance can be guaranteed, the computing power of the terminal device is extremely high, and the recognition efficiency cannot be guaranteed

Images