Time sequence data automatic labeling method and device, electronic equipment, medium and product

Abstract

The invention relates to a time sequence data automatic labeling method and device, electronic equipment, a medium and a product, and the method comprises the steps: obtaining to-be-labeled frames, arranging the to-be-labeled frames according to a time sequence, inputting the to-be-labeled frames into a target detection model, and outputting a detection box frame by frame as a corresponding first automatic labeling result; correcting a corresponding first automatic labeling result based on first frame data in the to-be-labeled frame and a preset detection frame area value and position; based on to-be-labeled frames arranged according to a time sequence, the corrected and correct first automatic labeling result is input into the target tracking model, and all detection frames of the object in the subsequent frames in the frame data are predicted to serve as second automatic labeling results of the corresponding frames; and comparing the detection frame of each object in the second automatic labeling result with the detection frame of the first automatic labeling result, and updating the first automatic labeling result of the corresponding frame. The method has the effect of providing the labeling result interacting with the labeling process.

Description

technical field [0001] The present application relates to the technical field of data labeling, and in particular, to a method, apparatus, electronic device, medium and product for automatic labeling of time series data. Background technique [0002] At present, deep learning models often require a large amount of labeling data, and labelers need to perform a lot of labeling work, and providing accurate automatic labeling results of objects can greatly reduce the workload of labelers. [0003] However, the existing automatic labeling method provides the object labeling results based on the original data before the labeling starts, without combining the real-time labeling information generated during the labeling process, the labeling results of the objects cannot be adjusted and corrected in time during the actual labeling process, and automatic labeling cannot be corrected immediately. The result is wrong, resulting in inaccurate automatic object annotation results provided...

AI Technical Summary

Problems solved by technology

[0003] However, the existing automatic labeling method is to provide object labeling results based on the original data before the labeling starts, without combining the real-time labeling information generated during the labeling process, the object labeling results cannot be adjusted and corrected in time during the actual labeling process, and the automatic labeling cannot be corrected immediately The result is wrong, resulting in inaccurate automatic object labeling results

[0004] In view of the related technologies mentioned above, the inventor believes that the existing automatic labeling method has the defect that the object labeling results cannot be adjusted and corrected in time during the actual labeling process, resulting in inaccurate automatic object labeling results provided

Images