masakiplusReferring to existing illustrations helps novice drawers to realize their ideas. In order to find such helpful references from a large image collection, we first build a semantic vector representation of illustrations by training convolutional neural networks.
2015/09/15 リンク