CPC G06V 10/25 (2022.01) [G06T 3/4046 (2013.01); G06T 7/50 (2017.01); G06T 7/73 (2017.01); G06V 10/462 (2022.01); G06V 10/82 (2022.01); G06V 40/11 (2022.01); G06T 2207/10028 (2013.01); G06T 2207/20084 (2013.01); G06V 10/803 (2022.01)] | 19 Claims |
1. A hand pose estimation method using an image feature extraction network and a predetermined depth classification network, the method applied in an electronic device and comprising:
obtaining, by a ROIAlign feature extractor, a feature map corresponding to a hand depth image;
inputting the feature map into the image feature extraction network to obtain an image information set feature map corresponding to the hand depth image;
up-sampling, by an up-sampler, the image information set feature map to obtain a target resolution feature map;
inputting the target resolution feature map into the predetermined depth classification network to obtain depth maps corresponding to hand key points in the hand depth image, wherein the predetermined depth classification network is configured to distinguish hand key points of different depths;
determining, based on the depth maps, depth values corresponding to the hand key points; and
performing, based on the determined depth values corresponding to the hand key point, hand pose estimation.
|