Deep learning skin disease picture comparison and classification method, storage medium and robot

Abstract

The invention discloses a deep learning skin disease picture comparison and classification method, a storage medium and a robot. The invention aims to solve the technical problems that when a classification algorithm for extracting features is adopted to classify images, the classification algorithm for manually extracting the features consumes long time, and triage efficiency is reduced, The method comprising the following steps: S100, classing skin disease pictures into training pictures; S200, performing data enhancement processing on all the training pictures; S300, inputting the trainingpictures in a picture database into a convolutional neural network, and carrying out continuous training; S400, optimizing the convolutional neural network according to the weight loss calculated by aloss function, and stopping training and fixing the weight after the optimal convolutional neural network is obtained; and S500, inputting to-be-classified patient pictures into the optimal convolutional neural network model, and classifying to obtain a picture classification result. The effects of automatically training to classify the pictures and greatly improving the picture classification efficiency to improve the triage rate are achieved.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a deep learning method for comparing and classifying skin disease pictures, a storage medium and a robot. Background technique [0002] Dermatosis is a general term for diseases that occur on the skin and skin appendages. There are many types of skin diseases, and common skin diseases include: viral, bacterial, fungal and allergic skin diseases. The identification and process qualitative analysis of skin diseases is a huge workload. [0003] The existing Chinese patent with the publication number CN107247958A discloses a method for extracting skin disease features based on image recognition. The method includes the following steps: firstly, preprocessing the sample skin disease pictures acquired, and then vertically segmenting the images and Carry out the corresponding geometric transformation, on this basis, extract the feature area of ​​different types of skin diseas...

AI Technical Summary

Problems solved by technology

[0004] The above existing technical solutions have the following defects: when using the classification algorithm for extracting features to classify images, professional personnel are required to intervene to set the features, and because it is difficult to extract high-quality features, the classification algorithm for manually extracting features It takes a long time and reduces the efficiency of triage, so it needs to be improved

Images