A method, device and system for establishing a machine learning classification model

Abstract

The present application provides a method, device and system for establishing a machine learning classification model, wherein the method includes: detecting a label frame containing a product body in a product image; The product body is segmented to obtain a first segmentation result; the first segmentation result and the pre-marked second segmentation result are used as samples, and the samples are used for deep learning training to obtain a classification model. By adopting these technical solutions provided in the present application, the workload of manually labeling samples can be reduced, and the practicability of the classification model can be enhanced.

Description

technical field [0001] The present application relates to the field of machine learning, in particular to a method, device and system for establishing a machine learning classification model. Background technique [0002] In recent years, image segmentation methods have played a very important role in the development of the Internet. Because the display of images on the Internet requires a comprehensive consideration of aesthetics and click-through rate, especially on e-commerce websites. Therefore, website designers need to carry out adaptive design for the subject in the image, such as adding text material, color adjustment and so on. The subject in the image is the part that characterizes the product itself, which is separated from the original image by the image segmentation method. [0003] At present, there are two types of commonly used image segmentation methods, one is the traditional segmentation method, and the other is the segmentation method based on deep lear...

AI Technical Summary

Problems solved by technology

This type of classifier only uses simple, artificially designed color and texture features to segment. Due to the variety of shooting environments, shooting angles, and product types, the classification model is not effective in segmenting product images on the Internet. ideal

[0005] Among them, the segmentation method based on deep learning uses pixel-level image samples to learn a classification model through machine learning, but the deep learning process requires a large number of manually labeled pixel-level image samples. In addition, the later optimization of the classification model also relies on manual labeling Pixel-level image samples, a large amount of labor costs limit the practicability of deep learning-based segmentation methods

Images