hand key points on the original image

[CURRYZJW]

Hello, thank you very much for your contribution, please how can I get the coordinates of the corresponding hand key points on the original image ...

[CURRYZJW]

2D skeleton

joint_img_xy1 = np.concatenate((joint_img[:,:2], np.ones_like(joint_img[:,:1])),1) joint_img = np.dot(bb2img_trans, joint_img_xy1.transpose(1,0)).transpose(1,0)

I found that the generated hand key point information is not consistent with the coordinates of the pixels in the original image, what should I do?

[mks0601]

https://github.com/facebookresearch/InterWild/blob/1a0da99386f60fad7cba802d0188fcf6b9fff57f/demo/demo.py#L157 joint_img is in the original image space

[CURRYZJW]

Thanks, can you provide a little bit of the location of the distribution of the 21 keys in one hand.Like, which one is 0

[CURRYZJW]

Is that so?

[mks0601]

https://github.com/facebookresearch/InterWild/blob/1a0da99386f60fad7cba802d0188fcf6b9fff57f/common/utils/mano.py#L38 lower number means closer to the finger root

[CURRYZJW]

Thank you very much for your reply