search

mdbloice / Augmentor

Image augmentation library in Python for machine learning.
https://augmentor.readthedocs.io/en/stable
MIT License
5.08k stars 866 forks source link

Transforming labels accordingly #169

Open fmbahrt opened 5 years ago

fmbahrt commented 5 years ago

Let's say my labels are landmarks on an image, am I able to transform them aswell - such that they fit on the transformed image?

mdbloice commented 5 years ago

Hi @fmbahrt no currently this is not supported. I am assuming the landmark labels are x,y coordinates on the image?

fmbahrt commented 5 years ago

@mdbloice yes, the landmarks are (x, y) coordinates.

mdbloice commented 5 years ago

@fmbahrt OK understood, no unfortunately that is not currently supported... M.

jay-reynolds commented 5 years ago

I also have a problem that requires having the transformation (generally, operations) used to generate the augmented result and am interested in helping out here.

mdbloice commented 5 years ago

Hi @jay-reynolds, sorry for the delay replying, yes sure, by all means make a pull request. It's a feature quite a few people have asked for so it would be worthwhile doing I am sure.