An image recognition method based on improved ArcFace loss function

Abstract

The invention discloses an image recognition method based on an improved ArcFace loss function, the feature of image is extracted by using the image recognition network based on depth learning, the extracted features are used to describe the main information of the image, the improved ArcFace loss function is used to train the image recognition network based on depth learning. The improved ArcFaceloss function maximizes the classification boundary by decreasing the intra-class distance and increasing the inter-class distance in the angle space, so as to improve the accuracy of image recognition model. The invention is used in the field of pattern recognition.

Description

technical field [0001] The invention belongs to the field of deep learning for extracting image features by a deep neural network, relates to technologies such as neural networks and pattern recognition, and in particular relates to an image recognition method based on an improved ArcFace loss function. Background technique [0002] With the advent of the era of big data and the substantial improvement of computing power, image recognition technology is developing towards advanced semantic understanding, and image recognition technology based on deep learning has become a research hotspot in the field of artificial intelligence today. [0003] Image recognition technology is a technology that automatically processes, analyzes and understands images through computers to identify targets and objects in various patterns. Image recognition technology has a wide range of applications. For example, in agriculture, it can be judged by the growth of plants, the color of leaves and f...

AI Technical Summary

Problems solved by technology

Image recognition technology has a wide range of applications. For example, in agriculture, it can be judged by the growth of plants, the color of leaves and flowers, and water, fertilize, and kill insects for plants; in industry, image recognition can be used to monitor the entire workshop In medicine, the health status of patients can be analyzed through the shape of cells and bones; in aerospace, aerospace research can be carried out based on real-time feedback of satellite images; in daily life, image recognition technology is also very Popularity, such as license plate recognition, fingerprint recognition, etc.; however, there are still some difficulties in image recognition technology, due to the difficulty of image recognition due to changes in viewpoint, complex background, light and shadow changes, occlusion, deformation, etc., resulting in deep learning-based image recognition network training In the process of inaccurate image classification, in order to solve this problem, the ArcFace loss function was proposed, but ArcFace only maximizes the classification boundary by reducing the intra-class distance

Images