Use box center as reference point everywhere - #23
Conversation
| x: parseFloat(detectionOfThisFrameArray[2]) + w / 2, | ||
| y: parseFloat(detectionOfThisFrameArray[3]) + h / 2, |
There was a problem hiding this comment.
Shouldn't we subtract the width and height instead of adding it? Or am I misunderstanding the coordinate system?
There was a problem hiding this comment.
It should be correct that way. The format of the MOT17 input file is
<frame>, <id>, <bb_left>, <bb_top>, <bb_width>, <bb_height>, <conf>, <x>, <y>, <z>
so we're going from top left corner to center point, which means adding half of width and height.
|
This looks good to me but we will have to wait for @tdurand as I can't merge code in this repo :-) |
|
now you do @vsaw , feel free to merge, thanks for the PR.. sorry I don't have the bandwith to review / process this week |
|
@vsaw send me your npmjs.com account by email so you can release a new version on the package once merged.. |
|
I tested it locally. Looks good (see attached video) output.mp4Will merge and release a new version of the tracker thank you @akretz for finding and fixing this bug! |
3 participants
this fixes opendatacam/opendatacam#410