Image content article positioning method and device, equipment, medium and product

Abstract

The invention discloses an image content article positioning method and device, equipment, a medium and a product, and the method comprises the steps: obtaining a feature map of a target image, the feature map is composed of a plurality of plane points, and each plane point is a high-dimensional vector; the similarity between every two plane points is calculated, a similar state matrix is constructed according to the similarities, and elements in the similar state matrix represent that the plane points corresponding to row coordinates of the elements are similar or dissimilar to the plane points corresponding to column coordinates of the elements; pre-selecting a plurality of plane points to form a preliminary set according to the number of the similar plane points from small to large, and searching the plane points similar to the seed plane points with the least similar plane points to form a seed set; and counting the sum value of the similarities of each plane point in the feature map and all plane points in the seed set, and when the sum value is greater than a preset threshold value, judging that the corresponding plane point belongs to the article area. According to the invention, effective positioning of the content article in the target image can be realized, and the method has wide versatility.

Description

technical field [0001] The present application relates to the technical field of image recognition, and in particular to a method for locating an image content item, a corresponding device, computer equipment, a computer-readable storage medium, and a computer program product. Background technique [0002] Image positioning technology is a basic work in image recognition technology. In reality, a large number of application scenarios require the use of image positioning technology, including face recognition, object recognition, etc. By identifying the content object from the picture, the corresponding image of the content object can be cut out, and other images of the content object can be further processed. Work such as image retrieval, matching, labeling, etc. In the context of the rapid development of artificial intelligence technology, it can also be used to provide accurate representations of content object pictures for various neural network models for training corre...

AI Technical Summary

Problems solved by technology

[0004] The current item positioning technology requires a lot of manual labeling to train the positioning model. On the one hand, considering the large number of product categories in e-commerce, it is unrealistic to perform manual positioning and labeling of products in each category, and the corresponding cost Very big

On the other hand, unlike face images, which have obvious common characteristics, various commodity images in the e-commerce field have different shapes, and commodity images are often affected by sharpness, shooting angle, object size, background noise, etc. uneven quality

[0005] Therefore, it can be seen that various reasons lead to the poor migration ability and versatility of the existing widely used positioning models, and it is impossible to effectively locate unknown commodities.

Images