Considering recent developments in gene manipulation methods for honey bees, establishing simple and robust assay systems which can analyze behavioral components in detail inside a laboratory is important for the rise of behavioral genetics in the honey bee. We focused on the antennal movements of the honey bee and developed an experimental system for analyzing the antennal responses (ARs) of the honey bee using DeepLabCut, a markerless posture-tracking tool using deep learning. The tracking of antennal movements using DeepLabCut during the presentation of vertical (downward and upward) motion stimuli successfully detected the direction-specific ARs in the transverse plane, which has been reported in the previous studies where bees tilted their antennae in the direction opposite to the motion stimuli. In addition, we found that honey bees also exhibited direction-specific ARs in the coronal plane in response to horizontal (forward and backward) motion stimuli. Furthermore, an investigation of the developmental maturation of honey bee ARs showed that ARs to motion stimuli were not detected in bees immediately after emergence but became detectable through post-emergence development in an experience-independent manner. Finally, unsupervised clustering analysis using multidimensional data created by processing tracking data using DeepLabCut classified antennal movements into different clusters, suggesting that data-driven behavioral classification can apply to AR paradigms. In summary, our results revealed direction-specific ARs even in the coronal plane to horizontal motion stimuli and developmental maturation of ARs for the first time, and suggest the efficacy of data-driven analysis for behavioral classification in behavioral studies of the honey bee.