A Difficult Sample Image Screening Method and Device for a Video Object Detection Model

Abstract

The embodiment of the present invention discloses a method and device for screening difficult sample images of a video target detection model. The method includes: detecting whether the current video frame image of the surrounding environment collected by the acquisition device in real time is received; if so, the frame number interval between the current video frame image and the last video frame image for full image target detection is preset At intervals, perform full-image target detection on the current video frame image; when the position and category of the detected target are detected and the local target detection is performed on the previous video frame image to obtain the position and category of the detected target, when the previous video frame When there is a first target that does not match the detected target of the current video frame image among the detected targets of the image, determine that the current video frame image is a difficult sample image and store it, and return to perform detection whether the surrounding environment collected by the acquisition device in real time is received The steps of the current video frame image. Applying the solutions provided by the embodiments of the present invention can reduce the screening cost of difficult sample images.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a method and device for screening difficult sample images of a video target detection model. Background technique [0002] The training of the video object detection model depends on large-scale sample images, and the general sample images cannot improve the performance of the video object detection model. Therefore, at present, the performance of the video object detection model is mainly improved through difficult sample images. [0003] At present, there are many screening methods for difficult sample images, mainly including threshold screening methods, supervised learning classifier methods, and anomaly detection methods. The above three methods are all for screening each frame of image collected by the acquisition device, resulting in a large amount of calculation. It makes the screening cost of difficult sample images higher. Contents of the invention [0004] T...

AI Technical Summary

Problems solved by technology

[0002] The training of the video target detection model depends on large-scale sample images. General sample images cannot improve the performance of the video target detection model. Therefore, currently, difficult sample images are mainly used to improve the performance of the video target detection model.

[0003] At present, there are many screening methods for difficult sample images, mainly including threshold screening methods, supervised learning classifier methods, and anomaly detection methods. The above three methods are all for screening each frame of image collected by the acquisition device, resulting in a large amount of calculation. Makes the screening of difficult sample images more expensive

Images