Include toe keypoint - Body-Foot Estimation

[SoCohesive]

Hello! Is there any way to include the foot (toe keypoints) in your model or are they excluded for a reason? I noticed in the OpenPose sample here: https://github.com/CMU-Perceptual-Computing-Lab/openpose

There is a toe keypoint but I am unsure how to convert that model to CoreML. If it can be included in your current model that would be amazing!

@tucan9389 any insight on this would be greatly appreciated !

[tucan9389]

@SoCohesive Sorry for late. I think there are many ways. But If you want to use my repo, please check this repository. You can customize the model architecture (like add new Keypoint) and train with own dataset which must be annotated foot keypoint.

[SoCohesive]

@tucan9389 thank you so much for that. So I have the new dataset and understand how to add the json

Can you explain a bit more on how to customize the model architecture?

[tucan9389]

I didn't customize the architecture of PoseEstimationForMobile but just train with my own fingertip dataset (annotated). You can annotate with annotation tools. In my case, I annotated my own dataset using tucan9389/KeypointAnnotation tool.

[SoCohesive]

Hi @tucan9389 how can I modify the model structure here - https://github.com/edvardHua/PoseEstimationForMobile

To return 4 keypoints instead of 14 with different names? Do I modify the mv2_cpm.cfg file?

[tucan9389]

Not only change cfg file but also other python script file 6 month ago. I don’t know the current structure of the project now.

[yubeiluo]

Hi @tucan9389, first, I want to say thanks for your work. In my recent work, I need all the 25 key points for pose estimation, per you discussions above, do I need to annotate the missing keypoints on my own?

[tucan9389]

@yubeiluo I'm sorry for late reply.

Hi @tucan9389, first, I want to say thanks for your work. In my recent work, I need all the 25 key points for pose estimation, per you discussions above, do I need to annotate the missing keypoints on my own?

Yes. But If you find the dataset annotated 25 keypoints what you want, no need to annotate again.