Hi Jon,
You can always check the latest changes in our
changelog. But in version 6 we did a deep change in our identification algorithm, its details will be explained in an article to be released very soon. Until then I can try to respond to your questions.
Knowledge transfer is a way to reuse the weights of a previously trained network as a starting point for the new one. It's used to improve tracking times. Since the networks we use in v6 are different from the ones in v5, knowledge transfer from v5 can sometimes be ignored by idtrackerai (depending on the difficulty of the video to track) but, in any case, it should be handled appropriately by the software and the weights should be used if needed.
Identity transfer (which means keeping the animal labels from one tracked video into a new one) should be fine from version 5 to 6.
And, if you already tracked all your videos without identity transfer, you can still match identities between these by using the included tool
idmatcherai. It's ready to handle trackings from both version 5 and 6.
If you encounter any error while working with any of these options, please report it to our email
in...@idtracker.ai.
Hope this helps your research,
Jordi