Skip to content
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

how you get the '1.25' #46

Open
laoxihongshi opened this issue Mar 10, 2018 · 2 comments
Open

how you get the '1.25' #46

laoxihongshi opened this issue Mar 10, 2018 · 2 comments

Comments

@laoxihongshi
Copy link

https://github.com/umich-vl/pose-hg-train/blob/2fef6915fbd836a5d218a5d2f0c87c463532f1a6/src/util/dataset/mpii.lua#L94

thanks for your wonderful works!
I want know how you get the 1.25?
when i change this another number, the training become hard.

@anewell
Copy link
Collaborator

anewell commented Mar 12, 2018

To get 1.25, I manually looked at the output of generateSample in util/pose.lua to check that the majority of crops included the full body of the target figure while also ensuring that they filled the image as well as possible. If you change the value, it is likely that the figure will be too large in which case the head and ankles will get cropped out, or too small in which case the resolution is lower and the precision of the predictions will drop.

@laoxihongshi
Copy link
Author

thanks for your help!!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants