tensorflow for segNet on Jetson Nano

Autonomous Machines Jetson & Embedded Systems Jetson Nano
1

in jetson-inference, there is sample for segnet, but it supports only caffemode.
do we have some sample for segnet which support tensorflow.
Thanks

2

Hi,

We only have a caffe-based segnet model and sample.

If you already have a TensorFlow model, it’s recommended to update sample to support uff model.
Here is tutorial of uff model for your reference:
[url]https://docs.nvidia.com/deeplearning/sdk/tensorrt-developer-guide/index.html#import_model_c[/url]

Thanks.

3

Hi AastaLLL,

Thanks for your reply.
I have flow the tutorial to convert my tensorflow model to uff model,
and try to load and parse the uff model, it is not very convenient.

4

Hi,

Sorry that we need an intermediate model format for better maintenance.
You can always report here if meeting any issue or error.

Thanks.

5

Hi AastaLLL,

I have try to load the uff mode, and encounter this problem:

ERROR: Parameter check failed at: …/builder/Network.cpp::addInput::406, condition: isValidDims(dims)
ERROR: UFFParser: Failed to parseInput for node Placeholder
ERROR: UFFParser: Parser error: Placeholder: Failed to parse node - Invalid Tensor found at node Placeholder
ERROR: sample_uff_mnist: Fail to parse
ERROR: sample_uff_mnist: Model load failed

I want to find the source file: Network.cpp, but i can’t find it in tensorRT folder.

waiting for your help.
Thanks very much.

6

Hi AastaLLL,

I have resolved this error. The rootcause is that the dims in my model is not match with the parameters in the code.

The last problem:
It costs about 5min in this function:
ICudaEngine* engine = builder->buildCudaEngine(*network);

it is too long, if only happens at the first time, that is ok, but it happens every time.
How can i resolve this problem.

Thanks very much.

7

Hi,

Good to know this.
Thanks for your update. : )

8

Hi, I met the same problem. Have you solved it? #ERROR: Parameter check failed at: …/builder/Network.cpp::addInput::406, condition: isValidDims(dims)