Predicting Drosophila Body Orientation from a Translational Trajectory using an Artificial Neural Network

A fully connected feedforward ANN predicts Drosophila body yaw orientation from translational flight trajectories with a median error of 5.0 degrees.

• Why it matters: Body yaw is important for analyzing insect flight, but it is hard to measure over full trajectories without specialized hardware, limiting large or long experiments.
• What they did: The ANN was trained on freely flying Drosophila data where both trajectory and body yaw were recorded, using ground and air velocity vectors as input features.
• The result: The model predicts yaw with a median 5.0° error and substantially outperforms linear and Mobius regressions during crosswind or downwind motion, enabling yaw recovery from center-of-mass-only datasets.

AI summary of the abstract · may contain errors · not medical advice · read the paper.