Training Network with sub networks

Question

I am planning to train the following network from end-to-end. I have two questions: Question 1

I have 4 ground truths.

1. Segmentation of distorted images
2. Parameters
3. Corrected images
4. Segmentation of corrected images

My problem is that as it can be seen in the image, The first two outputs are from network 1 and 2 and the last 2 from network 3.

Is there any way I can train my network from end to end.

Question 2

How can I load the VGG network weights. I have connections from middle layers to Network 2 and subnetwork 1-2.

I am using Tensorflow 2.4.1