The golden-collared manakin (Manacus vitellinus) is a small tropical bird, which lives in the Panama forest. The males perform elaborate, acrobatic displays to court mates [l]. During its courtship dance the male demonstrates its physical strength by jumping between saplings, producing loud wing snaps mid-flight. Mating success seems to be related to superior motor skills [2], which allow the male to execute its dance faster and more precisely. However, it is not fully clear yet how exactly the courtship dance has to be performed to impress a female. To gain more knowledge about their dance, biologists recorded the birds in the wild with high-speed cameras at 60 fps. One of the videos can be found at 1. Manually annotating the male bird in every frame of the videos to enable analyzing their behavior is a tedious process. We propose a novel approach for automatic visual tracking of the male golden-collared manakin, combining a convolutional neural network, background subtraction and a Kalman filter.