Deep Visual-Semantic Alignments for Generating Image Descriptionscs.stanford.edu4 pointspeterkchen12 years ago