predictive-filter-flow
predictive-filter-flow copied to clipboard
aimerykong
Code for pose tracking in mgPFF paper
Hi, do you have the code for pose tracking? I wonder how exactly do you encode the 15 keypoints as 15 maps?
Hi,
Thanks for your interest. I would release that after our submission becomes accepted. But here is how we did that --
- prepare 15 maps for 15 keypoints respectively.
- for each map, I dilate the corresponding joint to make it bigger in size, thus allowing for flowing; otherwise it would diminish.
- using the PFF to propagate the 15 maps, forming a new stack of 15 maps.
- for each map, voting for the new location of the joint.
Hope this helps understand the workflow.
Regards, Shu
On Wed, Apr 24, 2019 at 10:34 AM zlai0 [email protected] wrote:
Hi, do you have the code for pose tracking? I wonder how exactly do you encode the 15 keypoints as 15 maps?
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/aimerykong/predictive-filter-flow/issues/2, or mute the thread https://github.com/notifications/unsubscribe-auth/ABRJSJEK5AYUGA6OO6XUV3TPSCKZLANCNFSM4HIGMC7A .
Hi Shu,
Sure, thanks for your quick reply. I will try that. But I was just wondering how do you dilate the joint, and how is the voting done?
I just simply used scipy.ndimage.binary_dilation to dilate each joint before propagating; then as for voting, I compute the coordinate mean (based on mass center) of the the propagated dots.
Regards, Shu
On Wed, Apr 24, 2019 at 11:47 AM zlai0 [email protected] wrote:
Hi Shu,
Sure, thanks for your quick reply. I will try that. But I was just wondering how do you dilate the joint, and how is the voting done?
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/aimerykong/predictive-filter-flow/issues/2#issuecomment-486378543, or mute the thread https://github.com/notifications/unsubscribe-auth/ABRJSJBBHXWHRTTNTCAYNUDPSCTKRANCNFSM4HIGMC7A .