Michael de Gans

See #64 for an idea of how this might be accomplished. Work will need to be done on the config parser class to make the suggested changes, and the camera...

So, one issue with .ini files is that Sections can't have duplicate keys (at least without modifying the parser), so I just solved this in the deepstream branch by having...

So, multiple standard sources are implemented in the Gstreamer branch and I'm currently paused waiting on the pipeline end and distancing code. @mhejrati Any objection to me continuing where I...

So, one thing I noticed trying to solve this last night in C is that distance calculations may be unnecessary. So, I haven't closely looked at your distance caculation code,...

@alpha-carinae29 Aha. But the bounding box height changes with distance too, doesn't it? If a person is far away, the box height is ~six feet in real world units. If...

Hey, so it works better than I thought actually. Many thanks for Nvidia and the recent CUDA course that i took recently that gave me the idea. If I hadn't...

So here's a video of the above code on Xavier. For some reason the bboxes are green, while on x86 they are red. This appears to be a bug in...

Here it is on x86, with the proper color bounding boxes ~except for a small bug with the channel value clipping fixed. When some values go above 1.0, the box...

> Nice. @mdegans I am a bit confused now with this thread and the other on #50, is there any problem with using this code for post-processing? So, the above...

Hey, so I noticed [non_max_suppression_fast](https://github.com/mdegans/neuralet/blob/master/applications/smart-distancing/libs/core.py#L157) is copypasta from [this blog](https://www.pyimagesearch.com/2015/02/16/faster-non-maximum-suppression-python/) with the comments removed. I think we'll need a license to use it unless pyimagesearch has some page somewhere with...