Bar Plot from Scientific Research

The smart-Kage behavioral labeling accuracy (top) is evaluated as the ability of the classifier to avoid false and find true positives (F1 score). The number of ground-truth frames used to calculate the phenotyping performance is shown at the bottom; drum exploration (L-dr, R-dr) is a much sparser behavior compared to running on the wheel (W) or quiescence states (Q). Each T-maze trial prediction was immediately checked against the ground-truth activation of IR sensors in the nose-poke ports, resulting in 100% accuracy in this category (not shown).

DOI