Deploy Tensorflow trained ssd_mobilenet in jetson nano

Autonomous Machines Jetson & Embedded Systems Jetson Nano
,
1

Hi there, I finally managed to train a network in tensorflow and I want to use it in the nano. I know how to use a pretained one and a the ones trained on the nano with pytorch. But I don’t know how to call the network that I exported from my pc into a python script in the nano. It was a ssd mobilenet.Please guide me.

3

Hi,

You can install TensorFlow for Nano with the instructions below:

After installation, you can use the same inference python script on Nano.

Thanks.

1 Like
4

Thanks for your response, I fallowed this repo to train networks in my PC using a 2070 super How to Train Your Own Object Detector Using TensorFlow Object Detection API - neptune.ai and then y exported it as showed here TensorFlow Object Detection API: Best Practices to Training, Evaluation & Deployment - neptune.ai this gonverted a pb file to onnx. I hoped to be able to use the exported onnx file as I use the ones exported by the nano here jetson-inference/pytorch-ssd.md at master · dusty-nv/jetson-inference · GitHub. But the onnx file from my pc do not run in my nano. I am new at programming and AI so there is a lot that escapes my comprehension. Could you recomend me a way to get a file from my pc runing in my nano. Prease provideme with some code example that I can write in the nano script to get it working.
For example tu run the jetson onnx I write net2=jetson.inference.detectNet(‘ssd-mobilenet-v2’,[‘–model=/media/kc/1.0T/4th/jetson-inference/python/training/detection/ssd/models/fruit2/ssd-mobilenet.onnx’,‘–input-blob=input_0’,‘–labels=/media/kc/1.0T/4th/jetson-inference/python/training/detection/ssd/models/fruit2/labels.txt’,‘–output-cvg=scores’,‘–output-bbox=boxes’,‘–camera=/dev/video0’],threshold=.7)
. Thanks for your time and help.

5

Hi,

jetson-inference use TensorRT as a backend inference engine.
You can deploy it with TensorRT to validate if it is workable with TensorRT or not.

/usr/src/tensorrt/bin/trtexec --onnx=[your/onnx/model]

Thanks.

1 Like
7

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.