Seal identification from drone images

Hello seal people and WildMe team,

I have recently discovered Seals Wildbook, and I feel it will be super useful for my grey seal research in the Baltic Sea. So far, I’ve been successfully using it to identify females captured through a remote camera on one of our breeding sites, but now I’m also starting to try it out on drone images. Does anyone have any experience with individual identification from aerial photos? Anything to have in mind? Should it work well with the current algorithms, or do they have to be trained differently?

I have loads of drone images from the breeding sites, and I’ve been cropping the females that are lying on their side, with clear/semi-clear patterns. The images are very small and the quality is quite poor, but when I uploaded a couple of females who I already knew by eye, Hotspotter managed to identify them correctly - so that’s a good sign!

1 Like

Great question! I’m not personally aware of other drone images used in Seal Wildbook, but I know that there are researchers on Flukebook who use drone images to identify cetaceans. Maybe they can offer advice on best practices. @jwaite

If aerial images are provided in training data when a new species is requested, they will be more likely to be readily matched. However, MiewID in particular is good at matching viewpoints it hasn’t been previously trained on, though I understand MiewID doesn’t always perform as well as Hotspotter does with seals.

I’m curious to hear about the results of your continued tests with matching from aerial images!