Research Press Release

Animal tracking by visual fingerprints

Nature Methods

June 2, 2014

A new method to generate unique visual fingerprints for single animals filmed in a group is reported in a paper this week in Nature Methods. This method will enable researchers to reliably recognize and track and re-identify individuals between sightings and across different videos, allowing robust tracking in challenging situations without the need to physically tag animals with unique identifiers.

Humans are good at distinguishing among other humans and following their movements and behavior, but they are poor at doing the same for individual mice, fish, flies or other animals. This makes the study of animals in groups of the same species very difficult.

Gonzalo de Polavieja and colleagues developed fully automatic software that computes a unique visual fingerprint-based on pixel values and their distances in isolated animal images-for each animal in a video recording of an animal group. They then used these visual fingerprints to identify and track the individual ‘fingerprinted’ animals in their groups under challenging conditions that cause existing methods to fail.

DOI:10.1038/nmeth.2994 | Original article

Research highlights

PrivacyMark System