parser gives weird results

[trekkete]

So i'm not using deepstream but I have the resnet10 model and I'm trying to parse the results. I tried adapting the parser found in nvdsinfer but for a frame of 1280x720 i get coordinates like -30000 -5500 3200 45000.
I'm sure I have both cov and bbox, of dimensions (1, 4, 17, 30) and (1, 16, 17, 30).
Is there a guide to understand the various calculations? Maybe I interpreted some variables in the wrong way.
I know it's been a while since the last commit, and I thank you in advance for your time.