Backbone network for RetinaNet

[husseina-nvidia]

No description provided.

[tkornuta-nvidia]

Hey, this is for sure a step in the right direction. However, there are still some improvements required, as follows:

1. Full merge will involve using a script similar to the one that I have created, where:

• there is a NeMo DataLayer object that feeds the batches to the model,
• there is a Model that accepts those batches,
• we use NeMo training.

2. What I see now that forward() (which NeMo uses) is empty and you are calling some execute method, which is... I do not even understand what it is doing. Spanning many threads? What for? Where is forward step of the model?
3. The wrapper is not "compatible" from I/O ports definition, please take a look at the FasterRCNN create_ports()

Finally, please add retinanet-examples package as a dependency to setup.py. How do you install it? Can it be installed through pip, right?