Please tell me how to set the input for classification with the keypoint inferred from ncnn movenet

[manadopeee]

detail | 详细描述 | 詳細な説明

this is my code.

void NanoDet::detect_pose_cls(std::vector &points) { ncnn::Mat in_keypoints = ncnn::Mat ???; ncnn::Extractor ex_cla = poseCla.create_extractor(); // ex_cla.set_light_mode(true); // ex_cla.set_num_threads(4); ex_cla.input("input", in_keypoints); // ex_cla.input("data", in_keypoints); ncnn::Mat score, output; ex_cla.extract('score', score); ex_cla.extract('output', output); }

Perhaps the shape of the input points is 17, 2. It seems to consist of x and y coordinates per key point, and you can also calculate the score through points[0].score, points[1].score.

I want to input this in 17,3 shape, how do I do it?

Also, please tell me how to return the classification result as output and score.

thank you!!