-
Notifications
You must be signed in to change notification settings - Fork 77
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
root recover #8
Comments
Hi @moranli19 , |
@sameeroor @moranli19 Line 58 in 0808a4d
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
How to generate xyz_root from uv_root on evaluation set?
I have noticed that the xyz_root can be recovered from uv_root using the functions in your code (https://github.com/lmb-freiburg/freihand/blob/master/utils/model.py#L58) with 'scale' parameters which are parts of mano parameters[60]. It's obvious that such 'scale' is different from the bone length of each sample.
I want to know the exact meaning of 'scale' (mano parameter[60]). And the way to calculate xyz_root from uv_root on the evaluation set ('scale' parameter is not provided).
It seems that the root joint precision influences the final results (kp_mean_error (aligned)) on Codalab when using the Latent25D method[1]. So, it is important to generate a precise root joint for each evaluation sample. Or, the alignment on Codalab should remove the influence of xyz_root.
Thanks for any reply!!!
[1]: Iqbal et.al. Hand Pose Estimation via Latent 2.5D Heatmap Regression.
The text was updated successfully, but these errors were encountered: