You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The shape and gray-level models that we train should really be their own classes. That way we could right built-in visualization functions that display the models in a more convenient manner.
Hi there--
Yes, you can train your own models on shapes other than faces. I can't guarantee the model will work well as a detector, but the code will let you use your own set of landmarks.
Thanks for your replying. it really helps me a lot.
And i have one more question that bothers me:do training set images and test set images have to be the same size?
I'm not sure, but I don't think the dimensions of the images within the train and test sets have to be equal. What matters is that the number of landmarks across both is identical.
The shape and gray-level models that we train should really be their own classes. That way we could right built-in visualization functions that display the models in a more convenient manner.