CrowdPet

Automatic identification of animals in images

SciPet SciPet and FAPESP PIPE, phases 1 and 2 2016-2023
Lead researcher
Unicamp Inventors Award 2019, ‘Licensed Technology’ category
1st place Unicamp Ventures Entrepreneur Award 2018, ‘Social Impact case’ category
Finalist INOVA Unicamp Award 2017, as Victor Capone advisor
A world that is controlled by algorithms
Smartphone app uses computer vision to identify animals
Where’s my Puppy? Digital Lost and Found

CrowdPet addresses the problem of identifying animals in images, also known as animal visual biometrics. The problem is: given a photo or photos of an animal, an automatic search should be made in the photos of previously registered animals (profiles) and return the profile of that animal that possibly is the one being searched, thus identifying it. One innovative aspect to consider is the open-set identification problem. A robust identification method should be able to point out when an individual could not be identified, i.e., when it was not previously registered. CrowdPet is a SciPet project and has received funding from FAPESP through the PIPE program, phases 1 and 2.