この資料をSMS送信: Self-supervised learning of structural representations of visual objects