Tao Toolkit Roc curve

I am trying to run the example of DetectNet-V2 and i want to get the ROC curve , accuracy and the confusion matrix of it how do i get it

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks

During detectnet_v2 training, you can get mAP, each class’s AP. You can find it in the training log, or status.json under result folder. Also, you can set tensorboard as well.
But currently there is not ROC curve, accuracy yet.

You can leverage tao_tensorflow1_backend/nvidia_tao_tf1/cv/detectnet_v2/evaluation/compute_metrics.py at 2ec95cbbe0d74d6a180ea6e989f64d2d97d97712 · NVIDIA/tao_tensorflow1_backend · GitHub to get the groundtruth array, tao_tensorflow1_backend/nvidia_tao_tf1/cv/detectnet_v2/evaluation/compute_metrics.py at 2ec95cbbe0d74d6a180ea6e989f64d2d97d97712 · NVIDIA/tao_tensorflow1_backend · GitHub to get the prediction array.

You can docker run --runtime=nvidia -it --rm nvcr.io/nvidia/tao/tao-toolkit:5.0.0-tf1.15.5 /bin/bash. Then modify the code to meet your requirements.

vim /usr/local/lib/python3.8/dist-packages/nvidia_tao_tf1/cv/detectnet_v2/evaluation/compute_metrics.py`

More info can be found in How to calculate TPR and FPR in Python without using sklearn? | Ai Online Course.

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.