The iMet Collection 2019 Challenge Dataset
Abstract
Fine grained visual recognition has been a very popular research area in computer vision. With the rapid progress of deep neural networks in the recent decade, computer vision algorithms have been empowered to learn much more sophisticated representations for very complex semantics embedded in sub-domains such as natural world species and garment attributes. Computer vision technologies in artwork recognition, however, focuses more on instance retrieval or coarse-grained level in the past. In this work, we present a dataset and a public challenge on fine-grained artwork attribute recognition to leverage the wisdom from the community by further bridging the gap between modern artwork recognizing and computer vision technologies.