Display same side of individual in match results

What Wildbook should this feature be in? Flukebook

What would you like to see? For species being matched by dorsal fin (and/or saddle patch) such as killer whales, when showing match results by individual, display the best matched photo of the same side of the animal (if available). E.g. if the photo being matched is of the left side, show a left side photo of the possible matching individual.

How would this functionality help you? I suppose the computer may be able to mirror the image to assess the fin shape match, but it’s not as easy for a human verifier. So, displaying images of the same side will make the match verification easier. Also, when the match results show both L & R sides it is unclear whether the correct side was automatically annotated, or if the opposite side is being shown only because the wrong side was detected.


I’ll add to this also that for killer whales the saddle patch (which is not symmetrical from one side of the body to the other) is a very big component of how analysist confirm matches. When the preview for the proposed matches shows the wrong side of the animal, this would make the task of confirming matching much more tedious, requiring the analysist to go to the marked individual page for each proposed match to manually find and view a same side photo of that individual. In our use case of a small population of 74 animals, unfortunately the present display format (often showing opposite sides for proposed matches) means the time savings from the automation of proposed matches would be essentially offset by these extra steps, so that manually matching outside of flukebook would likely be more efficient.

That’s what’s supposed to happen. If right sides appear as matches when you’ve uploaded a left side, then one of those images was annotated with the incorrect viewpoint.

Hi Anastasia, good to know that’s the intended functionality. I’ll have to take a look at that and see if that’s what is going on. Many wrong side photos were being shown, so if that’s the case it would mean the annotations were automatically wrong a lot of the time, and seems odd that the wrong side of an animal would come up as a top match.

hi @Anastasia - I’ve had a chance to dig into this a bit more and have a few follow up questions.

Here’s one example of a case where the side shown is a mismatch.

The annotation for the photo being matched (on the left) just says IA class is ‘whale_orca’ without an indication of the viewpoint/whether this photo has been correctly classified as right side or not. The top 4 matches are all right side, which would seem to suggest it has, but then the 5th match (shown) and some others are of left sides. I found the annotation for this left side photo of J19 and it is classified as viewpoint = left. Do you have an idea of what might be going on here?

Also, if I want to check whether the annotation being displayed as a possible has the correct viewpoint assigned, what would be the best way to do so? In this case, I went to My Encounters, filtered by individual, sorted by sighting ID and then had to open each encounter for that sighting ID to find the one with the correct photo. Is there a more efficient way to search for the annotation/media asset ID from the matching page?

I should mention, I believe this annotation was a case where the viewpoint had been miss-assigned, so I had manually removed and reannotated the image (specifying right side) and resubmitted for matching, so I am not sure why this viewpoint is not being reported.

Also, if I do need to manually correct the viewpoint for an animal that already has an ID assigned, do I need to take any further steps (e.g. send to matching) after creating the annotation to ensure it will be considered for potential matches going forward?

I’m waiting on clarification from my teammates about the mixed viewpoints in match results and why the viewpoint description shows up on some annotations but not others.

Using your screenshot as a guide, I opened the encounter to match #5 putting my mouse in the green highlighted area in the space between J19 and the inspect button. That took me to the encounter this image came from.

It depends on your goal. If you want to see updated match candidates for that annotation with the new viewpoint/IA class you should re-run the match. If you just want to make sure it will appear in match results for future uploads, you don’t have to re-run it. The system will automatically look to match against the appropriate viewpoints/IA classes.

Sounds good, thanks.

great, that’s much more efficient. It would be nice if I could ‘ctrl click’ or right click to open it in a new tab without leaving the match results page (to save having to reload the page if I want to look at additional matches), but that doesn’t seem to be an option.

also great, thanks

Hard agree! I don’t know how likely the team is to implement this anytime soon, but I commiserate nonetheless.

1 Like

worth me submitting a feature request for, or is there one already that I can endorse with a heart/like?

I’m going to put that in our bug list because there’s some inconsistency in pages where you can right-click to open links in a new tab and others that either force you to open links in the same screen or force you to view the new tab and lose your place in the table of links you clicked from.

1 Like

Regarding the missing viewpoint when you hover over an annotation, it looks to be a bug. I’ve also included this in our bug list.

1 Like

Thanks! Do you know if there is a different way to view/confirm the viewpoint when it is missing?

The only other way I can think of from the user side is to do an Encounter Search and select the viewpoint from the Image Label Filters category, your username from the Metadata filters category, and then try and add other search info to increase the chance of turning up your specific encounter.

Hi Anastasia,

Following up to say I was doing some more checks and it seems like the match results does sometimes show matches for the wrong side of the animal, even when the viewpoint is correctly specified. So there is something going on beyond what you specified:

In the example below, the image being matched, on the left, correctly has viewpoint = right. The proposed match on the right, also correctly has viewpoint = left, yet is being shown as a match result.

Thanks for following up! I’ll add it to our bug list, but I wonder if the fact that these alternate viewpoints only turn up around the top 5-12 matches and not the top 4 results could mean that it’s trying to fill the top 12 spots when no more candidates for the correct viewpoints are available. As usual, I’ll follow up if I learn more.

By the way, if you can post a link to that encounter that’ll help us research it.

Sure, here it is.

For this case the #1 match is also of the mismatching side (while having the correct viewpoint labeled) so I don’t think that’s the case.

I appreciate it!