| CPC G06T 7/251 (2017.01) [G06V 10/255 (2022.01); G06V 10/766 (2022.01); G06V 10/7715 (2022.01); G06V 10/82 (2022.01); G06V 40/113 (2022.01); G06V 40/167 (2022.01); G06V 10/454 (2022.01); G06V 10/62 (2022.01); G06V 20/64 (2022.01)] | 18 Claims |
|
1. An object detection method, comprising:
obtaining a plurality of candidate bounding boxes of an interest object in a current image frame; and
filtering the plurality of candidate bounding boxes based on a determined bounding box of the interest object in a previous image frame, to obtain a determined bounding box of the interest object in the current image frame,
wherein obtaining the plurality of candidate bounding boxes of the interest object in the current image frame, comprises:
detecting the current image frame and obtaining a backbone feature map of the interest object in the current image frame;
obtaining a regression map based on the backbone feature map; and
determining N candidate bounding boxes based on the regression map, wherein Nis an integer greater than 1;
wherein the regression map comprises a bounding box confidence map, N location offset maps, and N shape offset maps; each of the N location offset maps indicates a location offset between a center of each candidate bounding box and a center of a ground truth box corresponding to the interest object; and each of the N shape offset maps indicates a shape offset between each candidate bounding box and the ground truth box.
|