Method and device for screening difficult sample data

Abstract

The embodiment of the invention discloses a method and device for screening difficult sample data. The method comprises the steps of performing visual feature extraction on each piece of to-be-screened data by utilizing a pre-established target detection model, and determining at least one suspected target region of each piece of to-be-screened data and a confidence coefficient corresponding to the suspected target region; determining a to-be-screened area from the suspected target area based on the first confidence threshold and the corresponding confidence; inputting the at least one to-be-screened area into a pre-established data screening model, and determining the data type of each to-be-screened area, the pre-established data screening model being a model obtained by training based on the data calibrated as the normal value type and the data calibrated as the abnormal value type; and determining the to-be-screened data containing at least one to-be-screened area determined as theabnormal value type as the difficult sample data, thereby improving the screening flexibility of the difficult sample data and improving the accuracy and recall ratio of the difficult sample data.

Description

technical field [0001] The present invention relates to the technical field of data screening, in particular to a screening method and device for difficult sample data. Background technique [0002] When training object detection models, the role of difficult sample data is often greater than that of easy sample data. In the training process of the target detection model, even a large amount of simple sample data is difficult to greatly improve the prediction accuracy of the target detection model, while difficult sample data often brings a large increase in the prediction accuracy of the target detection model. improvement. Then, how to screen out difficult sample data from a large number of sample data is very important. [0003] In related technologies, the process of screening difficult sample data from a large number of sample data is generally as follows: Using the current target detection model, from each sample data, determine at least one suspected target area and...

AI Technical Summary

Problems solved by technology

The process of screening difficult sample data is not flexible enough, and the accuracy and recall rate of the determined difficult sample data are not high enough

Images