How to Retrain an image Classifier for New Categories

Hello,
I’m using kit Nvidia Jetson TX2 and I trying retrain an image classifier with samples code retrai.py.
THis is my document: https://www.tensorflow.org/hub/tutorials/image_retraining

But I installed python2.7 and can not install tf-enviroment or annaconda.
Can you give me somethings about resolve to retrain the samples image with this kit.
Thanks for you support!
wait for you!

Hi lengoc.trung, have you tried installing the TensorFlow for Jetson TX2 package from here?

https://devtalk.nvidia.com/default/topic/1038957/jetson-tx2/tensorflow-for-jetson-tx2-/
https://docs.nvidia.com/deeplearning/dgx/install-tf-jetsontx2/index.html

yes, I’m installed.
But I don’t understand how to train an image with Tensorflow on JetPack environments.

Hi,

It’s recommended to get a desktop GPU for training rather than using Jetson.

Currently, we only release official TensorFlow package which targets for inference.
Some package you mentioned (ex. annaconda) may not work on aarch64 environment.

Thanks.

Hello AastaLLL,
Thanks for your response.
So I can’t training image on Jetson? I must training it on my desktop PC.
I can use Jetson for inference. So how to inference an image on google with file: retrain.py.
Thank you.

Hi,

It’s not recommended to apply training on Jetson due to embedded hardware limitation.

retrain.py is used for re-training.
If you just want to do the inference, please check other inference-based TensorFlow script.
Ex. https://github.com/tensorflow/hub/blob/master/examples/colab/action_recognition_with_tf_hub.ipynb

Thanks.

Hello,
THanks for your support.
But I trying re-training image with retrain.py.
My issue is run files retrain fails.