How to run tensorflow mobile net V2 model on Jetson Nano

Dear Sir,

I have trained my own Tensorflow SSD Mobile V2 model for num classes = 2

I have following files with me:-

frozen_inference_graph.pb

I wish to run this on jetson nano for inference.
How can I run it with Tensor RT ?

I am using Jetpack 4.3
Latest Version of Tensorflow on jetson nano
tensorrt == 6.0.1.10

Please tell me how can we do it.
Please provide me any link to code. I am so much confused after reading the blogs.

@AastaLLL @dusty_nv

Thanks in advance

Hi,

Please check this GitHub for information:

You may need to update the config.py into this version since TensorFlow has some update:

Then, the class number information can be customized as the following:

# config.py
...
NMS = gs.create_plugin_node(name="NMS", op="NMS_TRT",
    ...
    numClasses=3,
    ...

Thanks.

Thank you for your reply @AastaLLL

I am following the steps mentioned in your Github Repo, I am getting error as

OSError: libnvinfer.so.5: cannot open shared object file: No such file or directory

Please find screenshot attached herewith.

Hi,

This is because the sample is built on previous software release.
Please check this comment for the update of TensorRT 7.1.

Thanks.

Which comment ?
What do I need to do ?

Hi,

Sorry for the missing.
If you are using TensorRT 6.0, please check this comment for information:

Here is the information for TensorRT 7.1:

Thanks.