Hi.
How to find the accuracy value for validation and inference?
Do the object detection models support tensorboard?
When you run evaluation commands, for example,
tlt detectnet_v2 evaluate
It will show the accuracy.
When you run inference commands, for example,
tlt detectnet_2 inference
It will automatically generate bbox rendered images in output_path/images_annotated . To get the bbox labels in KITTI format, configure the bbox_handler_config spec file using the kitti_dump parameter as mentioned here. This will generate the output in output_path/labels .
You can compare the labels with the groudtruth labels.
Currently, there is not tensorboard.
During training, end users can also find the accuracy result for the specified epochs.
I use yolov3, yolov4 and retinanet and only see train loss, validation loss, average precision and mAP during training.
And just average precision and mAP during validation
Yes, that’s the expected behavior.
See some result matrix in https://developer.nvidia.com/blog/preparing-state-of-the-art-models-for-classification-and-object-detection-with-tlt/
and https://developer.nvidia.com/blog/creating-a-real-time-license-plate-detection-and-recognition-app
yes, but it does not include accuracy
Usually the object detection networks show the precision. For classification network, it will show accuracy.
In TLT classification network, the accuracy is available in the result.
1 Like
This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.