Getting Classes from Training Test

[eanmikale]

Dusty, During training, it would be extremely useful if we can see which classes are underperforming during testing, so we may more efficiently adjust data-sets between training. Is there a way to do this with the jetson.utils library? Thank you.

[dusty-nv]

Hi @eanmikale, that wouldn't be a jetson.inference/jetson.utils feature, because those libraries are used for runtime inferencing only. This would be something to be added to the PyTorch training scripts. In theory it's possible, but I haven't looked into it. I get the PyTorch training scripts from other sources.

[eanmikale]

For anyone else who might be interested, it seems this might be the closest we can gethttps://stackoverflow.com/questions/62958248/calculate-accuracy-for-each-class-using-cnn-and-pytorch. We can close it as far as I am concerned. Thanks Dusty.  ---- On Wed, 23 Feb 2022 12:48:36 -0600 @.*** wrote ---- Hi @eanmikale, that wouldn't be a jetson.inference/jetson.utils feature, because those libraries are used for runtime inferencing only. This would be something to be added to the PyTorch training scripts. In theory it's possible, but I haven't looked into it. I get the PyTorch training scripts from other sources.

—Reply to this email directly, view it on GitHub, or unsubscribe.Triage notifications on the go with GitHub Mobile for iOS or Android. You are receiving this because you were mentioned.Message ID: @.***>