Matthew Dawkins

Given the comment about not finding groundtruth, svm training probably doesn't support -it /tmp/tmp8f377cwf/input_truth_list.txt option yet, will see how hard it is to add

Training directly off of videos will be supported soon, but it's not now. You'd need to extract the frames from video (e.g. using the extract frames script in the project...

Seeing build errors with this on windows, also the branch shouldn't contain large swaths of commented code if going into master/next

I changed this branch's target to be master now, but there are conflicts and unresolved issues (breaking outside of a virtual env).

I cherry-picked the python3 stuff from this branch, but did not merge it due to breaking things using native python outside of a few virtualenvs, also now needs a rebase.

The feature temporarily doesn't work unfortunately, but will be added back in sometime over the next month.

Just to confuse matters, but we're adding a new desktop GUI (a port of VIAME-Web just running locally) in the next few days that has better polygons and detection/track attributes

I have no objections, just the time cost of setting it up and deciding what the copyright banner should say

It's in the works, the easiest way to tell in the current implementation is to open up the log files in database/Logs and see what frame # it's currently on...

I think a Spanish translation is a great idea, though the main issue I see is that our English documentation is still being updated at a fast rate and they...