Overlapping detection boxes in detect-netv2?

I am currently working on training of single class Person having dataset details below.

Training images : 21,156
Testing images : 5600

All my images are resized (320*240).

After training, when i tested model on deep-stream, it working well. Now the issue I am facing is that it show overlapping of two objects.
Below is some samples attached.
overlap_2

Reference: DetectNetV_2: bounding box issue

1 Like