Text this: Large-scale capture of hidden fluorescent labels for training generalizable markerless motion capture models.