Multiple inference on Jetson Nano

Robotics & Edge Computing Jetson & Embedded Systems Jetson Nano
1

Hi, can I run multiple inference from different model weight on Jetson Nano and how do I go about this

Thanks

2

Hi bvgohmslf, for multiple model weights, you will need to load multiple models and perform the inferencing independently on them.
This will be the same approach regardless of if you are using TensorRT or an ML framework like TensorFlow, PyTorch, ect.

The other approach would be to re-train the model to support classification of the multiple objects in the same model.

3

How can I do the later, kindly show me an example of re-training a model to support classification of the multiple objects in the same model. Thanks

4

Here’s an example using PyTorch that I’ve been working on that reshapes the classification network to support a different number of classes and re-trains it with transfer learning: https://github.com/dusty-nv/jetson-inference/blob/python/docs/pytorch-cat-dog.md

5

Thanks but isn’t there anyone that combines vision and voice. That’s where I am more interested in. Thanks

6

Do you mean image captioning? You could check out Neuraltalk2, although the last time I tried that was on TX2.

If you mean combining image classification and voice recognition, I personally don’t know of a model that combines both, probably easiest to run them independently in that case.