Inference Models compliant with TensorRT 10.3

[baqwas]

While stepping through the exercise at Classifying Images with ImageNet, I’ve encountered a roadblock - the links to the models in the article are not compliant with TensorRT 10.3. Since my knowledge is very limited regarding model access, I need help in locating suitable models that will help me to complete the exercise without the following error message (for most models listed in the article):

[TRT]    TensorRT 10.3 does not support legacy caffe models
[TRT]    device GPU, failed to load networks/ResNet-18/ResNet-18.caffemodel
[TRT]    failed to load networks/ResNet-18/ResNet-18.caffemodel
[TRT]    imageNet -- failed to initialize.

I built from source using the article at Building the Project from Source. Thanks for any guidance on how to leverage the Orin Nano for these types of image classification exercises.

Regards.

[AastaLLL]

Hi,

From v10.3, the Caffe-based model is not supported in TensorRT.
If you want to test the Caffe model with jetson-inference, please set up your device with JetPack 6.0.

Thanks.

[baqwas]

Hello @AastaLLL,

I am using JetPack 6 (I believe, :)). I simply want to do some tutorials with the TensorRT (current version is 10.3?). Where can I find suitable TensorRT compliant models, if any, at one of the Jetson Developer sites? Thanks.

Regards.
P.S.
Caffe presumably is a red herring for me since I didn’t even anything about it until the error message!

[AastaLLL]

Hi,

You can find some TensorRT samples in the below link:

C++:

TensorRT/samples at release/10.3 · NVIDIA/TensorRT

Python:

TensorRT/samples/python at release/10.3 · NVIDIA/TensorRT

Thanks.

[baqwas]

Great! Thanks again for all your support.

Regards.