Understanding parameters of training config

Hi Guys,

I am trying to understand what the parameters for “detectnet_v2” imply. I am having difficulty in the following parameters:

Postprocessing config:

a) minimum_bounding_box_height - Is this parameter applied in original image dimension or resized image dimension which is mentioned in augmentation - preprocessing config ?

evaluation_config:

a) evaluation_box_config - minimum_height, maximum_height, minimum_width, maximum_width. Are these parameters applied in original image dimension or resized image dimension which is mentioned in augmentation - preprocessing config ?

Please help me out.

Thanks.

Hi neophyte1,
For detectnet_v2, as previous mentioned, all of the images must be resized offline to the final training size and the corresponding bounding boxes must be scaled accordingly.

So, there are only resized images and bboxes for training. So, the parameter are applied to the resized image dimension.

More info:
minimum_height : Ground truths with height below this value are ignored.
maximum_height : Ground truths with height above this value are ignored.
minimum_width : Ground truths with width below this value are ignored.
maximum_width : Ground truths with width above this value are ignored.