Seol mar théacs é seo: Deep convolutional neural networks for efficient pose estimation in gesture videos