Text this: Human pose estimation in image sequences