After checking back this morning, it looks like one bulk import successfully completed all detections, but these two have not although they say detection is complete.
Detection did complete on both encounters (identification couldn’t start if it hadn’t). These images where the fins are very distant aren’t as well represented in training data where it was trained on clear left and right dorsal images.
You’ll need to manually annotate any missed detections and start a new match from there.
This would be good data to use for future retraining efforts once it’s been curated. The next time we’re ready to retrain for dolphin data, we may reach out to you.
Thanks, I was wondering that too so great to get a confirmation from you, and I’ll keep that in mind as I am processing images for import into Flukebook.
I would love to help with any future retraining, I have a lot of similar data that could be useful.