Romero-Ferrero, F. et al. Nat Method 16, 179–182 (2019)

Need to track one animal moving around among many? A new species-agnostic tool, idtracker.ai, from researchers at the Champalimaud Center for the Unknown in Lisbon combines two neural networks to track unmarked animals within a larger group. Each network follows an elongated “blob” that represents an individual animal in a recording. One algorithm keeps track of each blob while the second detects when blobs touch or cross one another. The results suggest the tool can accurately track individual zebrafish and fruit flies in groups of up to 100 animals, as well as medaka fish, ants, and mice in smaller groups. For best performance, the researchers found that at least 30 images with individual animals are needed to train the algorithm but that it can perform against different backgrounds and account for unusual behavior among the animals.