Networks for vision