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

关于bounding box问题 #15

Open
hao-qiang opened this issue Feb 20, 2019 · 2 comments
Open

关于bounding box问题 #15

hao-qiang opened this issue Feb 20, 2019 · 2 comments

Comments

@hao-qiang
Copy link

您好,您的方法十分高效准确,我正在复现您的工作,有两个小的细节想向您确认一下:

  1. 您实验中用到的bounding box是根据landmarks边界确定的吗?是否向外扩展了。
  2. 使用WFLW训练时,大概需要训练多少epochs,学习率如何设置。
    谢谢
@FengZhenhua
Copy link
Owner

@hao-qiang 你好,关于你的问题:

  1. 我们一般使用数据库提供的bbox值,但是由于WFLW的bbox有时会导致crop的区域不完整,所以我们把训练集的bbox向外扩展了25%,然后加上bbox抖动。
    2.在WFLW上进行训练,对CNN6和预训练的RESNET50,我们分别调整学习率从3e-3->3e-6和1e-3~1e-6,训练的epoch数量分别是800和200.

@hao-qiang
Copy link
Author

谢谢您的详细解答,我还想询问一下我的L1 loss设置和NME设置是否正确 (98个关键点),再次感谢。

xi_abs_diff = |xi_pred - xi_true|
yi_abs_diff = |yi_pred - yi_true|
L1 loss=(x1_abs_diff+x2_abs_diff +...+x98_abs_diff+y1_abs_diff+y2_abs_diff +...+y98_abs_diff) / (98*2)

mean_distance = [(xi_pred - xi_true)^2 + (yi_pred - yi_true)^2]^(1/2) / 98
NME = mean_distance / 双眼中心距离

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