Object boundary identification method, device and equipment

Abstract

The invention relates to an object boundary identification method, device and equipment, and the method comprises the steps: identifying a target object in an input image, and obtaining a target object image; filling the target object image into a square, and performing binary image segmentation on the target object image to obtain a binary image; performing full-image boundary processing on the input image, and determining a full-image boundary; and connecting intersection points obtained by intersection of the whole image boundary and the boundary of the binary image to obtain the boundary of the target object. According to the method, picture boundary prediction can be more accurate, operation is easy when secondary modification is conducted manually on the picture boundary prediction, and due to the fact that it can be guaranteed that the prediction result is more accurate, manual boundary compensation is facilitated through the inaccurate part.

Description

technical field [0001] The present application relates to the field of image processing, and in particular to an object boundary recognition method, device and equipment. Background technique [0002] With the widespread application of supervised learning, the demand for data is increasing, and image data annotation is a big gap. At present, many labeling companies mainly use the combination of preloading and manual labeling when processing image data labeling. However, due to the limitations of the algorithm and the poor matching of the algorithm processing results with manual secondary labeling, etc., The preload step often doesn't work to its full potential. [0003] At present, the more commonly used instance segmentation algorithms cannot meet the fineness requirements of fine labeling, so it is necessary to manually modify the algorithm recognition results twice. Because the existing algorithm can't control many details in place, it adds a lot of workload to the manu...

AI Technical Summary

Problems solved by technology

At present, many labeling companies mainly use the combination of preloading and manual labeling when processing image data labeling. However, due to the limitations of the algorithm and the poor matching of the algorithm processing results with manual secondary labeling, etc., The preload step often doesn't work to its full potential

[0003] At present, the more commonly used instance segmentation algorithms cannot meet the fineness requirements of fine labeling, so it is necessary to manually modify the algorithm recognition results twice

Because the existing algorithm can't control many details in place, it adds a lot of workload to the manual, and sometimes even consumes more time than not preloading

Images