Two things can happen when you see mismapped spots in a scan result.
a. The match results are stored in an XML file. However, the images they pull to map the matched spot pairs to are the latest extraction images. Therefore, when someone remapped the spots, your scan results did not change, but they're now mapped to the new extraction image that was uploaded. The best way to resolve this is:
b. Flash has a cache for images and patterns. This can also cause a spot mismap if you've looked at a scan result then remapped and rescanned. The second set of results might show the first mapping. In this case, clear your browser cache and then reload the page.
This situation can occur when the first submitted image/video is not of a supported image or video file type. This can be fixed by logging into Wildbook and uploading a supported image/video type OR by asking the webmaster (webmaster at whaleshark dot org)to resolve the bad file (if you don't have required access). If you upload a supported file type, follow the instructions below to regenerate the image thumbnail.
1. Log into Wildbook.
2. Go to the appropriate encounter page.
3. If you have edit permissions for the encounter, look in the Action/Edit bar for the Reset Thumbnail box. Choose the submitted image to generate the thumbnail from and then click Reset Thumbnail.
4. Go back to the page where you first encountered the “BAD FILE” thumbnail and Refresh/Reload the page in your browser. The updated thumbnail will appear.
Looking at the process backwards, the end goal of whale shark mark-recapture is to obtain a large group of sharks with low misidentification for accurate population modeling. Our photo-identification standards, as presented in this wiki, are in place to help support accurate matching across the data sets of multiple users…the assumption being that some of us might be sharing sharks and therefore we need standards to ensure that we process data identically and can quickly and accurately match them between our data sets when they appear. The pattern recognition algorithms provide a reliable, fast way to do that. Ultimately we could get the same effect (very laboriously and less accurately) with “by eye” matching, but this adds the risk of double counting a shark if not properly matched to an existing photo across catalogs. The pattern recognition software we use (Modified Groth and I3S algorithms) significantly reduce this risk, and therefore we use them as a standard for all new sharks (i.e. the patterns provide a measure of statistical evidence that there is not a match elsewhere in the catalogs).
That said, there will always be some human-added variability and error, including some missing/extra mapped spots and of course variable angle between shark and photographer. We know that our probability of automated matching degrades (for both the I3S and Modified Groth algorithms) when: *angle between photographer and shark flank falls away from perpendicular *spots outside the patterning area are added *too few spots from the patterning area are added
The algorithms can internally account for some of this variability (Spot! helps too), but we have to make a judgment call for each unmatched pattern and peer review it. The questions we ask are: *Are all of the proper spots mapped? *Is the pattern properly rotated? *Is the angle between photographer and shark appropriate, roughly within fifteen degrees of perpendicular on either side and with minimal roll?
Ultimately, it's a judgment call based on those criteria, and we have peer review and discussion (even sometimes disagreements) about whether a shark is new. We're simply trying to ensure that every new shark has the maximum probability of being matched in the future.