Sharkbook over-writing / misclassifying C. taurus IDs

In which Wildbook did the issue occur?


What operating system were you using? (eg. MacOS 10.15.3)

MacOS Sonoma 14.0

What web browser were you using? (eg. Chrome 79)

Brave 1.57.62

What is your role on the site? (admin, researcher, etc)


What happened?

Encounter is here: Sharkbook: Wildbook for Sharks

Okay, so this is an odd one – I’m uploading an archival dataset, and the first photo has a ‘main photo’ and an ‘inset photo’ of the same shark. I spot-mapped the ‘main photo’ first, and sent it for matching, then processed the inset photo (which are both left-side IDs). However, the second left side has overwritten the first image, and the scans are labelled as a Right and Left.

What are some steps we could take to reproduce the issue?

Please check the encounter link to see the issue, as it’s unusual.

However, I’m concerned that the spot-mapped image was incorrectly classed as a Right Side, as this follows a different image (processed by Anastasia) also being incorrectly labelled:

(It’s labelled two right sides as a left and right side)

If this is a bulk import report, send the spreadsheet to with the email subject line matching your bug report

That’s interesting. I wonder if spot mapping works on the assumption that there’s only one animal in each photo. This could explain why the previous feature region for the main image was overwritten.

I’m working on a separate guide for spot mapping Carcharias taurus since the whale shark one isn’t exactly a good 1:1 guide for it and I’m not 100% I mapped the dorsal fins correctly. Give me some more time to dig into this.

No worries at all – thanks!

I don’t suppose you know if I3S or Modified Groth typically produce better results for this species? I know that I3S was originally designed for C. taurus; the lists produced are quite different.

I don’t but I’ll find out and report back with what I find.

1 Like

Statistically speaking, we don’t know which algorithm is better for the species. For small amounts of matching, they may be the same. For larger data sets (thousand of patterns), we know from whale sharks that Modified Groth still matches accurately while I3S begins to lose accuracy. PIE v2 is taking over from Modified Groth for whale sharks. At tens of thousands of patterns, Modified Groth begins to lose accuracy and PIE is the best for whale sharks. This may or not be the progression path in the future for sand tiger sharks.

This was fun to research. Spot mapping allows only one left side pattern and one right side pattern per Encounter, no matter how many pictures or animals are in them. Looking back at your image, because both the main image and the inset are of the same side of the same shark, you’ll only be able to spot map one of these.

As mentioned above, if two spot maps are created for one encounter, it assumes it’s because there’s both a right and a left image included. I also learned that the anchor point for labeling the first and second dorsal is at the front of the dorsal fins, where the fin meets the body. As I suspected, I did not place them correctly in my attempt to spot map it, so we’ll want to delete that one and start again. One more thing I wanted to call out before you do:

You’ll want to make sure the blue line of the yellow box is level with the back of the shark. See how Spot-mapped image 2 is rotated compared to Spot-mapped image 1? Consistent positioning helps ensure you’re getting accurate match candidates to review.

Let me know if you want to walk through this in a call.

1 Like

Thanks so much for all of this great information! I’m re-processing both of those sharks (split the encounter with the inset photo into two encounters; started again with the right-side shot above).

1 Like

My pleasure! It’ll take me a while longer to get a video and help documentation completed, but I’ll let you know when that’s live.

1 Like