Noob with tensorflow/keras experience needs help implementing Detectron2

[yogeshriyat]

Hello,

I have a computer vision classification project that I built out in Tensorflow/Keras. I used data generators to read in the image files and output the results in a way that are readable for Keras models (in this case I use the model, Xception). Here is a link to my notebook: https://bit.ly/3FKPTHO

I'd like to implement Detectron2 and I am interested in panoptic segmentation. I did go through the tutorial, but it just wasn't super helpful for me as a starting point. Can someone guide me on the steps that I would need to take? Maybe I need to first learn about Pytorch's image generators? Where do I then find the model architecture to feed this input data into?

Maybe this tutorial can be customized/tweaked to work with Detectron2??
https://pytorch.org/tutorials/beginner/transfer_learning_tutorial.html

Any guidance would be much appreciated!