TensorRT Engine Inference

Autonomous Vehicles DRIVE PX 2 General
1

Hello everyone!

I am looking for some advice regarding Tensorrt inference. I want to run a Tensorflow model on the Drive PX2. The steps on how to convert a Tensorflow model into an TensorRT engine are clear to me, but I am not quite sure what to do next.
How can I initialize this engine and feed new data (e.g. camera as input) to it on the Drive PX2.

Thanks in advance!

2

Dear raphaDev,
As Tensorflow is not supported by driveworks currently, You can check to use EGLStream with NvMedia as producer and CUDA as consumer. Please refer Autonomous Vehicle Development Platforms | NVIDIA Docs. Once you have data in CUDA buffers, you can feed it to network as shown in any TensorRT sample.

3

Dear SivaRamaKrishna,

thank you for the answer! I will look into it and try to implement it in this way.