US 11,810,319 B2
Image detection method, device, storage medium and computer program product
Xiangbo Su, Beijing (CN); and Jian Wang, Beijing (CN)
Assigned to Beijing Baidu Netcom Science & Technology Co., Ltd, Beijing (CN)
Filed by Beijing Baidu Netcom Science and Technology Co., Ltd., Beijing (CN)
Filed on Jun. 9, 2021, as Appl. No. 17/303,896.
Claims priority of application No. 202011461590.X (CN), filed on Dec. 11, 2020.
Prior Publication US 2021/0295088 A1, Sep. 23, 2021
Int. Cl. G06T 7/62 (2017.01); G06T 7/73 (2017.01); G06T 7/70 (2017.01); G06V 10/46 (2022.01); G06V 10/48 (2022.01); G06V 10/25 (2022.01)
CPC G06T 7/73 (2017.01) [G06T 7/62 (2017.01); G06T 7/70 (2017.01); G06V 10/25 (2022.01); G06V 10/462 (2022.01); G06V 10/48 (2022.01); G06T 2207/20084 (2013.01)] 14 Claims
OG exemplary drawing
 
1. An image detection method, comprising:
performing target detection on a feature map of an image to be detected through a target branch network to obtain target detection information of candidate anchor boxes at a pixel position of the feature map;
performing keypoint detection on the feature map of the image to be detected through a keypoint branch network to obtain heatmap groups of the candidate anchor boxes at the pixel position of the feature map; and
obtaining positions of keypoints of a target object and information about the target object in the image to be detected according to the target detection information of the candidate anchor boxes and the heatmap groups of the candidate anchor boxes;
wherein obtaining the positions of the keypoints of the target object and the information about the target object in the image to be detected according to the target detection information of the candidate anchor boxes and the heatmap groups of the candidate anchor boxes comprises:
determining, according to the target detection information of the candidate anchor boxes, the information about the target object in the image to be detected, and using a candidate anchor box associated with the information about the target object as a target anchor box;
selecting a heatmap group of the target anchor box from the heatmap groups of the candidate anchor boxes; and
determining the positions of the keypoints of the target object according to the heatmap group of the target anchor box;
wherein selecting the heatmap group of the target anchor box from the heatmap groups of the candidate anchor boxes comprises:
obtaining a target index number according to a size of the feature map, an anchor box number and the pixel position of the feature map of the target anchor box; and
selecting a heatmap group associated with the target index number from the heatmap groups of the candidate anchor boxes as the heatmap group of the target anchor box.