How can I optimize Tensorflow codes using TensorRT?

Autonomous Machines Jetson & Embedded Systems Jetson Nano
, , ,
1

Hi,

We are using a Github tutorial to implement object tracking on Jetson Nano 2GB, but the performance is not that good. We thought of using TensorRT to increase the performance.

Any help to integrate our existing code to TensorRT engine would be great.

The repository that we are using for our code is - GitHub - theAIGuysCode/yolov4-deepsort: Object tracking implemented with YOLOv4, DeepSort, and TensorFlow.

Thanks

3

Hi,

For TensorFlow user, please convert your model into ONNX format.
And you can deploy it with TensorRT directly.

Thanks.

4

Hi,

Could you please share any reference or tutorial that we could refer, it would be of great help.

Thanks

5

Hi,

You can convert the model into ONNX with tf2onnx.
And the ONNX file can be deployed with trtexec directly.

$ /usr/src/tensorrt/bin/trtexec --onnx=[onnx/model]

Thanks.