Image recognition method and device, equipment and storage medium

Abstract

The embodiment of the invention discloses an image recognition method and device, equipment and a storage medium, and the method comprises the steps: obtaining a to-be-detected image, determining a target contour frame image corresponding to the to-be-detected image, and enabling the target contour frame image to comprise at least one target contour frame; inputting the target contour frame imageinto a target segmentation network model to obtain at least one target offset corresponding to each target contour frame output by the target segmentation network model, the target segmentation network model being obtained by training a pre-constructed to-be-trained segmentation network model; and adjusting the position of each corresponding contour point in the target contour frame according to each target offset to obtain a target image, thereby solving the problem of relatively low image recognition accuracy, obtaining a target segmentation network model by training a to-be-trained segmentation network model, accurately predicting the target offset corresponding to the target contour frame. And the position of each contour point is adjusted according to the target offset, so that an accurate target image is obtained, and the data processing precision is improved.

Description

technical field [0001] Embodiments of the present invention relate to the technical field of image processing, and in particular, to an image recognition method, device, equipment, and storage medium. Background technique [0002] Semantic segmentation of images obtains pixelated dense classification by judging which class each pixel in the image belongs to among the predefined categories. With the development of deep learning technology, convolutional neural network (CNN) has been more and more widely used in semantic segmentation. Especially since Long et al. first used fully convolutional networks to segment natural images end-to-end in 2014, semantic segmentation has made great breakthroughs. Semantic segmentation only semantically classifies each pixel, and cannot distinguish pixels belonging to the same category but different objects. For example, it can segment repeated adjacent continuous characters or overlapping multiple people as a whole, and cannot distinguish e...

AI Technical Summary

Problems solved by technology

Top-Down Based is also called the Two Stage method. In 2020, the latest open source algorithm DeepSnake was proposed by Zhejiang University. It is a Two Stage algorithm for instance segmentation. The deviation is determined according to the characteristics of a fixed number of adjacent contour points around the desired contour point. However, the offset is only determined based on the features of a fixed number of adjacent contour points around the desired contour point. The learned features have low accuracy and cannot meet the requirements of high-precision data processing, and cannot achieve accurate recognition of images.

Images