Unsupervised Learning For Object Representations By Watching And Moving