Howdy – I’ve uploaded a video (a few actually) to the VIAME web portal (username bsemmens), and I’m looking for some guidance for which tracker to start with. We would like to pull out as many fish as possible from high resolution video for instance pattern recognition work (facial recognition for natural tag mark recapture). Both the boxing and tracking are a very important part of this – tracking is important so we don’t produce hundreds of images/boxes of the same fish and ultimately swamp the instance pattern recognition workflow. Advice appreciated!!
Sorry for the delayed reply - there’s a couple existing trackers in the GUI/system, then a couple we’ve experimented with, contained in python scripts outside of it.
If you want a quick first-pass option, running the ‘Utilities->Link Detections Default’ pipeline will run a default tracker on detections on individual frames, linking them into tracks.
If you trained a detector (single frame) pipeline and want to permanently add a tracker to it this can be done by editing the configuration file for that pipeline, you can ping me with the trained detector name plus example data, and I can add one of two trackers or show you how to do it (matt.dawkins@kitware.com).