Image multi-task identification method and electronic equipment

Abstract

The invention provides an image multi-task identification method and electronic equipment, and the image multi-task identification method comprises the steps: inputting an image into a trained multi-task machine learning model; extracting first feature data from the image through a multi-task machine learning model, and determining a first recognition result according to the first feature data, wherein the first recognition result indicates the category of each pixel in the image; extracting second feature data from stacked images of the images and the first recognition result in the channel dimension through the multi-task machine learning model, and determining a second recognition result according to the second feature data, wherein the second recognition result indicates the type of the images; and outputting the first identification result and the second identification result by the multi-task machine learning model. The multi-task machine learning model not only has high-level comprehensive semantic analysis capability, but also has pixel-level recognition capability, and can accurately and credibly execute image multi-task recognition.

Description

technical field [0001] The invention relates to the technical field of image recognition, in particular to an image multi-task recognition method and electronic equipment. Background technique [0002] With the improvement of material living standards, more and more people rely on regular physical examinations to understand their own health status. Regular physical examinations can help people detect diseases in time to diagnose and treat diseases in time. However, regular physical examinations are accompanied by complex procedures and invasiveness. . Researchers have found that diseases such as hypertension, arteriosclerosis, and diabetes will show corresponding pathological changes on the retina, and the collection of fundus images has the advantages of non-invasiveness and convenience. Health) provides the possibility of high-frequency monitoring. [0003] At present, intelligent fundus image recognition technology based on deep learning has been widely used in health m...

AI Technical Summary

Problems solved by technology

It can be seen that only identifying the disease category may not be able to accurately locate the lesion, and it may not be possible to accurately determine the disease category only based on the identified lesion. The ability to identify tiny lesions, used to identify disease categories and corresponding lesions from fundus images

Images