Many different labs are currently exploring the possibilties of DeepLabCut annotated pose-estimation data. Even though, the possibilities of multi-animal tracking are providing an plethora of social and individual behavior to investigate, the interpretation of the DeepLabCut output, “X,Y” coordinates over time, remains complicated. The DeepOF module can calibrate the DeepLabCut data and export directly a set of individualistic- and social behaviors, using a supervised pipeline dependent on machine learning and rule-base annotations.
Set of supervised behavioral classifiers:
In addition, an unsupervised pipeline was employed using VQVAE autoencoder models.