Author response image 2. Triangulation error for each joint as a function of number of cameras used for calibration.

Mean absolute error (MAE) as a function of number of cameras used for triangulation and calibration, as well as human annotation versus network prediction. Calibration using all 7 cameras and human annotation is the ground truth (red circles). This ground truth is compared with using N cameras for triangulation and performing calibration from N cameras using either human annotation of images (blue circles) or network predictions of images (yellow circles). A comparison across joints (individual panels) demonstrates that certain joints (e.g., tarsus tips) are more susceptible to increased errors with fewer cameras.