Max has some preliminary results from training a machine learning model (deep network) to identify and track ants in videos with multiple ants. You can see that it’s pretty good at identifying each individual in the frame, can track most individuals over time (note that they are unmarked), except when multiple ants get too entangled and then it loses track of their identities. This is the first step towards learning to identify and classify interactions between ants. We’re mainly interested in characterizing aggressive interactions between non-nestmates, in the context of studies of the mechanism underlying nestmate recognition.