Automatic image annotation system and device oriented to unknown pre-training annotation data

Abstract

The invention discloses an automatic image annotation system oriented to unknown pre-training annotation data, and the system comprises an acquisition module, an image annotation module and a model migration module, wherein the model migration module is connected with the acquisition module and the image annotation module respectively; the acquisition module is used for acquiring a group of image annotation tasks and images to be annotated; the model migration module is used for migrating a group of determined pre-trained image processing models to an updated image processing model adaptive to a to-be-labeled image domain in an unsupervised manner, and comprises a model splitting unit, an information maximization loss constraint unit, a clustering unit, a label space classification unit, a label distribution unit, a distribution updating unit and a convergence unit which are connected in sequence. And the image annotation module is used for generating annotation information matched with the to-be-annotated image and performing visual annotation.

Description

technical field [0001] The present invention relates to the fields of artificial intelligence and computer vision, in particular to an image automatic labeling system and device which is oriented to pre-training labeling data agnostic. Background technique [0002] With the advent of the network information age, massive amounts of image data are generated every minute. For different vision tasks, training the corresponding deep learning models often requires a large amount of labeled data. However, the existing manual labeling method not only consumes a lot of manpower and material resources, but also inevitably has labeling errors. Therefore, various image automatic labeling methods have been proposed. At present, most automatic image labeling methods use supervised methods to train deep learning models to obtain high-accuracy labeling results. However, due to the influence of illumination, angle, background, etc. during image acquisition, there are large distribution di...

AI Technical Summary

Problems solved by technology

However, due to the influence of illumination, angle, background, etc. during image acquisition, there are large distribution differences between different data sets.

Therefore, it is difficult for such supervised methods to guarantee the generalization performance of the model on the data to be labeled

Images