Have you ever come in from the field with hundreds — or thousands — of dolphin fins that need to be processed for your catalogue?

The next time-consuming step typically involves cropping the fins from each photograph and saving each cropped image into your catalogue. Photo-id Ninja, originally titled Findr, was designed to automate that step: it used deep learning to extract the fins from the original image and returned the cropped image.

How it worked

It could not be simpler — drag your images into Photo-id Ninja, and a zip file came back containing the cropped images.

Stage II

Under development when the site was last updated. Photo-id Ninja was to go beyond detecting and cropping fins and help identify unique individuals across a set of images: upload all the photographs from a day in the field, and the cropped fins would be grouped by individual, making the comparison against a catalogue far quicker.

Who ran it

The project was a collaboration between researchers at Massey University’s Institute of Natural and Mathematical Sciences, the Coastal-Marine Research Group, and NIWA, New Zealand’s National Institute of Water and Atmospheric Research.

This site is an archived copy and is not operated by any of them.

Why dorsal fins

The trailing edge of a dolphin’s dorsal fin accumulates nicks, notches and scars through its life, and the resulting outline is distinctive enough to identify an individual and stable enough to recognise it years later. Photo-identification uses that: a catalogue of fins stands in for a register of animals, and it is built without tagging, handling or otherwise disturbing them — which is the whole reason the method is worth its considerable manual effort.