Ask about how to develop autonomous driving applications

Please provide the following info (check/uncheck the boxes after clicking “+ Create Topic”):
Software Version
DRIVE OS Linux 5.2.0
DRIVE OS Linux 5.2.0 and DriveWorks 3.5
[Yes] NVIDIA DRIVE™ Software 10.0 (Linux)
NVIDIA DRIVE™ Software 9.0 (Linux)
other DRIVE OS version

Target Operating System
[Yes] Linux

Hardware Platform
[Yes] NVIDIA DRIVE™ AGX Xavier DevKit (E3550)
NVIDIA DRIVE™ AGX Pegasus DevKit (E3550)

Hi, Community.

I am a new developer and encountered some questions during the development process. I see that there is a roadrunner program, but there is no source code.

So are there only two possible development methods, which of these two methods are generally used more? :

  1. Use drive works sdk to complete the development. By calling the API, use the Drivenet, LightNet and other models provided by DRIVE to complete the development of the application.

If these models are not suitable for local scenarios, is there any way to improve these models such as Drivenet?

  1. Completely develop the application by myself, including the model trained by myself.

If I train the model myself, do I need to match the model used by tensorrt to develop it?

Thank you

Dear @guaiyia,
Road runner is application built on top of DRIVEAV to show case capabilities. The source code of roadrunner is not publicly available.
You may use NV DNN modules in your use case if it helps. Note NV DNN models are not publicly available to fine tune. We are optimising the models in each DRIVE SW releaase. Also, If you want to deploy your custom DNN model into DW, you can use DW DNN module APIs. Please check DriveWorks SDK Reference: Samples. You can train your model first on any host and then need to use tensorRT_optimization tool to convert your model into DW supported model for deployment.