Deep Learning Enabled Image Feature Automatic Access System

In recent years,due to the technological breakthrough of artificial intelligence,deep learning,as an important branch of machine learning,has accelerated a new round of innovation process of image search technology.The working principle of image search is to extract the features of the target image,then extract the similar image from the image library by the way of image aggregation,a nd arrange the similar images according to the similarity from high to low.Therefore,it is necessary to update the image library frequently,that is,to update the images with missing features in the image library,so as to ensure the full coverage and high similarity of the images in the image library.In this paper,an automatic system for accessing image features is implemented.The main research results include :1.The system development task is divided into three subsystems: image scanning and writing subsystem,distributed image downloading subsystem and depth feature extraction subsystem.The image scanning and writing subsystem can scan and write image data to es server and image library efficiently and stably.The image download subsystem can download images on a large scale and preprocess them to generate the request of image feature extraction task.The depth feature extraction subsystem extracts the required features.2.The three subsystems form a data closed-loop as follows: scan the image data from ES,download it through the distributed image download system,and generate the feature extraction task request,then complete the feature extraction task by the feature extraction subsystem,and write it into es.3.In view of the key difficulties of the system,the original technology is proposed,such as automatic construction of feature extraction task.For the former,a method based on the principle of bitmap is proposed,while for the latter,a multi process and multi thread model based on Python is proposed to create a process cutting technology.4.Completed the development of the system,no need for human external intervention in the operation.The experimental results show that it not only reduces the labor cost,but also improves the accuracy and speed of image feature extraction.