Building a train/evaluation pipeline with Bounding Boxes

[jonieva]

Hi!

I'm new to MONAI and I'm trying to train a model that predicts bounding boxes (like RetinaNet, etc.).

I noticed there are a bunch of bounding box transforms implemented here, but I'm a bit confused on how to use them.

Could anybody point me to an example of how to apply transforms to images+bounding boxes during training, and the corresponding inverted transforms during inference?

Thank you in advance.

[KumoLiu]

Hi @jonieva, there is a folder for detection in the tutorial repo you could refer to.
Thanks!