How to show confusion matrix and model accuracy graph in Jetson Nano 2GB

How can I show the confusion matrix and model accuracy graph in Jetson Nano 2GB?

Hi,

Would you mind sharing more details about your use case?
Are you using our sample like jetson-inference or a third-party implementation?

Thanks.

I using jetson-inference

Hi @m012110024, jetson-inference is for inference with TensorRT and doesn’t include things like loading ground-truth datasets used for computing the confusion matrix or model accuracy. Those things would be more easily done with the code that was used for training the model (i.e. in the PyTorch or TensorFlow scripts)

Can you please provide me some example?

train.py used for image classification models already prints out the top-5 and top-1 accuracies. For the confusion matrix you would want to search for other examples of doing that in PyTorch (I also know that scikit-learn has convenient functions for computing the confusion matrix and other metrics, but can take awhile to compile for Jetson - although scikit-learn is already installed into the l4t-ml container)

Okay I also search about the scikit-learn too. Will try on it. Thank you so much