Is there a way to avoid getting warnings in TAO? More generally, does this software have options to choose what to output?
I am running a fasterRCNN model and I am getting hundreds of warnings like these:
WARNING: Got label marked as difficult(occlusion > 0), please set occlusion field in KITTI label to 0 and re-generate TFrecod dataset. if you want to include it in mAP Calculation during validation/evaluation
You can follow the warning info to set occlusion field in KITTI label to 0 and re-generate TFrecod dataset.
If you are not going to do this, you can also find the source code in https://github.com/NVIDIA/tao_tensorflow1_backend/blob/c7a3926ddddf3911842e057620bceb45bb5303cc/nvidia_tao_tf1/cv/faster_rcnn/utils/utils.py#L249-L255 and modify to your expected, and then replace it with /usr/local/lib/python3.8/dist-packages/nvidia_tao_tf1/cv/faster_rcnn/utils/utils.py .
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.