Automated analysis of bird head motion in unconstrained settings: A foundational study on semicircular canal evolution in archosaurs

This study presents a framework to automatically analyze head motion in birds from videos of natural behaviors. The process involves detecting birds, identifying key points on their heads, and tracking changes in their positions over time. Bird detection and key point extraction were trained on publicly available datasets, featuring videos and images of diverse bird species in uncontrolled settings. Initial challenges with complex video backgrounds causing misidentifications and inaccurate key points were addressed through validation, refinement, filtering, and smoothing. Head angular velocities and rotation frequencies were computed from the refined key points. The algorithm performed well at moderate speeds but was limited by the 30 Hz frame rate of most videos, which constrained measurable angular velocities and frequencies and caused motion blur, affecting key point detection. Our findings suggest that the framework may provide plausible estimates of head motion but also emphasize the importance of high frame rate videos in future research, including extensive comparisons against ground truth data, to fully characterize bird head movements. Importantly, this work is a foundational effort to understand the evolutionary drivers of the semicircular canals, the biosensor that monitors head rotations, for both extinct and extant tetrapods.

Methods
For the development of the bird head pose estimation (BHPE) module a new 2D BHPE annotated dataset is proposed, here entitled BirdGaze, which includes images from four prominent sources: the Animal Kingdom, NABirds, Birdsnap and eBird. These datasets represent the largest publicly available collections and are widely recognized in the literature for their significant role in avian research. Their extensive morphological diversity is crucial for this study. Besides the bird images, the proposed BirdGaze dataset includes a set of annotations, notably:

Center of the bounding box containing the bird body, described by its 2D coordinates;
Scale factor, defining a multiplying factor to apply to the bird bounding box for resizing it to fit a fixed rectangle size, which is used as input to the adopted key point extraction model;
Coordinates of the four selected 2D key points: top of head, tip of beak, left eye, right eye.