Learning Categorical Shape from Captioned Images | IEEE Conference Publication | IEEE Xplore