control label formatting for detected objects

[steveneighbour]

Noticed this today on a successful object scan, a person is picked up a decent distance from the camera (so they are pretty small) but the label drawn obscures the person.

Is it possible/feasible to control the label (eg remove the red background), in an effort to not obscure the detected object?

[blakeblackshear]

Looks like you are using 0.2.0-pre, right? What camera are you using? I need to tweak the position of the label a bit.

[steveneighbour]

i've been messing with the ffmpeg hardware decoding, so am using the latest git version from ffmpeg-subprocess branch.

These are from Hikvision cameras, at a pretty low res (width 640, height 360), 10fps

I'll have a quick play with util.py to alter the box, see if i can make it look cleaner

[steveneighbour]

i've added a car endpoint (just by copy/pasting your people endpoint), as that generates a lot more images than people in my street

Here are a few examples using different fonts/thickness/colours etc (i don't like any of them!)

I think perhaps stick with how you had it but, as you said, alter the positioning a bit so that small objects are not obscured so much?

Here is a car with the title removed entirely, i actually quite like that. I'm only really caring about cars and people, so including the label doesn't really add value for me.

Might try with a 1080p version as well, see if it looks better