computational models that predict which parts of an image the human eye will look at first based on pre-attentive features. You can approximate this yourself: squint at your video's first frame until everything blurs. The parts that still stand out are the parts with the highest salience. If your ke