Video-based instance tracking method and device, equipment and storage medium

Abstract

The invention discloses a video-based instance tracking method realized by adopting an artificial intelligence technology, and the method comprises the steps: obtaining a target feature map through a backbone network based on a target video frame in a to-be-detected video; according to the N instance bounding boxes, obtaining N bounding box regions of interest (ROI) from the target feature map; based on the N instance query vectors and the N bounding box ROIs, obtaining N first detection results through an instance segmentation network; determining at least one instance similarity according to the N first detection results and the M second detection results; and determining an instance tracking result of the target video frame according to the at least one instance similarity. The invention further provides a device, equipment and a storage medium. The end-to-end instance detection framework is constructed, instance detection independent of post-processing methods such as non-maximum suppression is realized, and the instance target is tracked based on the instance identifier, so that the instance tracking efficiency based on the video is improved.

Description

technical field [0001] The present application relates to the technical field of computer vision, and in particular to a video-based instance tracking method, device, equipment and storage medium. Background technique [0002] Instance segmentation is a crucial preprocessing for image recognition and computer vision, and is widely used in various fields. For example, instance segmentation can be used for tasks such as object recognition, object detection, and object tracking. In the object detection task, not only the category of the object in the image needs to be detected, but also the bounding box of the object needs to be detected. [0003] Currently, instance segmentation algorithms usually follow the process of "detect first, then segment", that is, detect and segment interesting instances in videos through object detection based on prior boxes. Specifically, when screening positive samples during training, it is necessary to perform matching based on the intersectio...

AI Technical Summary

Problems solved by technology

[0004] However, since the a priori box follows the principle of "one-to-many (that is, one real box corresponds to multiple a priori boxes)" during the training process, it is necessary to rely on the Non-Maximum Suppression (Non-Maximum Suppression) during the test phase. , NMS) and other post-processing methods to reduce repeated instance predictions, it is difficult to perform end-to-end reasoning, resulting in low instance tracking efficiency

Images