Why there are joint connections between [left_ear, left_shoulder] and [right_ear, right_shoulder]

michalfaber / tensorflow_Realtime_Multi-Person_Pose_Estimation

Multi-Person Pose Estimation project for Tensorflow 2.0 with a small and fast model based on MobilenetV3

Other

215 stars 65 forks source link

Why there are joint connections between [left_ear, left_shoulder] and [right_ear, right_shoulder] #15

Open hcv1027 opened 4 years ago

hcv1027 commented 4 years ago

As I read the code, I found there are two strange joint pairs in below code, [2, 16] and [5, 17], compare them with the image, it looks like that there are connection between [left_ear, left_shoulder] and [right_ear, right_shoulder]. I'm really confused about this part. May I ask if there is a reason?

joint_pairs = list(zip(
        [1, 8,  9,  1, 11, 12, 1, 2, 3,  2, 1, 5, 6,  5, 1,  0,  0, 14, 15],
        [8, 9, 10, 11, 12, 13, 2, 3, 4, 16, 5, 6, 7, 17, 0, 14, 15, 16, 17]))

keypoint_idx

mgarbade commented 3 years ago

As far as I know, this connection is part of the original CMU publication as well. Reason might be that it helps the model training. More paf connections might help finding both shoulder keypoints.