TF To TRT conversion

AI & Data Science Deep Learning (Training & Inference) TensorRT
1

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) Orin
• DeepStream Version 6.2
• JetPack Version (valid for Jetson only) 5.1
• TensorRT Version 8.5.2
I want to convert Facenet “.h5” keras model to TRT to be used by Deepstream. I have tried to follow ResNet-TF to TRT jupyter notebook available in this link Quick Start Guide :: NVIDIA Deep Learning TensorRT Documentation but too many attributes of resnet are not part of facenet model.
Is there any other documentation which i can follow for this conversion?

2

Hi,

This looks like a Jetson issue. Please refer to the below samples in case useful.

For any further assistance, we will move this post to to Jetson related forum.

Thanks!

4

the first link in not really related to my issue and the 2nd link shows torch2trt not tf2trt. Is there a clear documentation for tf2trt? My post has been already moved from jetson to TensorRT

6

Hi,

Which facenet are you using?
Is this one?

If yes, this is a TAO model and can be converted with TAO toolkit.
https://docs.nvidia.com/metropolis/deepstream/dev-guide/text/DS_TAO_integration.html

Thanks.

7

no facenet the face recognition model available in TF-KERAS

8

Hi,

For a custom model, please convert it into ONNX format and then generate the TensorRT engine with trtexec:

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

After that, you can use it with Deepstream by following the instructions below:
https://docs.nvidia.com/metropolis/deepstream/dev-guide/text/DS_using_custom_model.html

Thanks.