Can you describe what the issue is you’re experiencing?
I uploaded an encounter two weeks ago of a turtle (Pickle). Today I uploaded another picture of the same turtle from a different day, because I wanted to test if the algorithm will detect it as the same turtle. But it did not. However, this is crucial for the research my team is doing. We want to be able to upload encounters on a daily basis and find out if we encountered the same turtle on our reef on another day. We can not do it entirely manually as it is a lot of work, so we are hoping the algorithm can do it for us.
Can you provide steps on how to reproduce what you’re experiencing?
It looks like the image analysis pipeline is potentially looking for matches across all encounters of Eretmochelys imbricata instead of just the ones that match the location ID on the encounters. On your match page links, you can see it considered 11,595 candidates to determine a match. However, there are only about 64 encounters of this species in Madagascar. Too broad of a match filter can make correct matches appear lower in the match results or not display them at all.
I re-ran the match on the second, unidentified encounter (match results) and it correctly checked for matches only in Madagascar this time and found Pickle as the top candidate. This time it narrowed the candidate pool down to 18:
I’ll need more time to look into why IoT is looking for match candidates so broadly. Thanks for eltting us know. I’ll let you know as soon as I have an update. In the meantime, feel free to manually start any matches that seem questionable.
thank you so much for your quick response. I’ll try what you did with the narrowed candidate pool.
Also, should I run the matches with both algorithms or only the MiewID matcher. Does it make a difference?
