DetectNetV_2: bounding box issue

If you want to finetune the dbscan parameter, please run tlt-infer directly, set different dbscan in the tlt-infer spec file.

For example, refer to the command and spec file in my comment of the Retraining peoplenet model with own images - #14 by Morganh

Then, try to set a lower dbscan_eps, run tlt-infer and check the inference result.
dbscan_eps: (float) The search distance to group together boxes into a single cluster. The lesser the number, the more boxes are detected.

https://docs.nvidia.com/metropolis/TLT/tlt-getting-started-guide/text/creating_experiment_spec.html?highlight=dbscan#bbox-handler