Data Augmentation & Visualization example with Detectron2


I am able to get the Detectron2 work on custom dataset for instance segmentation, exactly following the Google Colab tutorial, by registering the custom dataset.

Using the Detectron2 framework -

  1. I would like to perform data augmentation on both images and annotations for MaskRCNN application. I would like the way of randomly selecting a transform from a list of transforms that PyTorch offers for data augmentation
  2. I also want to visualize some sample augmented images+annotations (as in Google Colab tutorial)

I read the detectron2 docs but I was not able to understand them.

Can anyone point me to some learning resources or reference code, so that I can understand and adapt it ?

Any help will be greatly appreciated :slight_smile:

The Data Augmentation docs give you an example how to use these transformations. Would that help? :slight_smile: