Local picture rapid detection method based on image content identification

Abstract

The invention discloses a local picture rapid detection method based on image content recognition, and the method mainly comprises the following steps: 1, building a training data set which comprisesthe collection of content text regions, central points and rotation angles corresponding to scene text pictures and annotation pictures in different noise environments; 2, training a text recognitionmodel and a target detection model at a server side by utilizing the training data constructed in the step 1; 3, automatically converting the image input of the client into a file with semantic annotation by using the trained text recognition model and the target detection model; 4, analyzing and arranging the files obtained in the step 3, and transferring the files into a local database of a user; and step 5, according to the request information sent by the client, performing retrieval in the database, then feeding back a retrieval result to the client, and displaying the picture content andthe corresponding label. According to the invention, the accuracy and speed of image retrieval can be improved.

Description

technical field [0001] The invention relates to machine learning and image processing, in particular to a fast local picture detection method based on image content recognition. Background technique [0002] In recent years, with the development of smart devices and the continuous improvement of camera pixels, people have gradually become accustomed to using cameras to record various information in daily life. But as more and more photos are accumulated in the album, how to quickly find the target picture in the album folder has become a new problem. [0003] Most of the existing search methods are based on file names, but album pictures are generally named according to the time or sequence of photos taken, and name-based searches are not suitable for this type of image query task. Manually renaming each picture and then retrieving it can complete the search task, but with the increase in the number of photos, the huge workload of labeling will undoubtedly bring a lot of ma...

AI Technical Summary

Problems solved by technology

[0003] Most of the existing search methods are based on file names, but album pictures are generally named according to the time or sequence of photos taken, and name-based searches are not suitable for such image query tasks

Manually renaming each picture and then retrieving it can complete the search task, but as the number of photos increases, the huge workload of labeling will undoubtedly bring a lot of manpower and time.

[0004] At present, most of the solutions to the above problems are based on fast renaming and image annotation. By optimizing the classification and annotation methods (such as batch renaming templates and smart label recommendations, etc.), although it can solve the problem that local images cannot be searched, However, the content of the album pictures is ever-changing, and different people have different naming habits. The above labeling methods will have great limitations in terms of accuracy and speed.

Images