Multi-feature fusion for fine-grained sketch-based image retrieval.

Fine-grained sketch-based image retrieval has become an important research topic in the computer vision area. To take advantage of more fine-grained information, we proposed a multi-feature fusion network for fine-grained sketch-based image retrieval. Multi-feature consists of a coarse-grained feature and two fine-grained features which can make better use of fine-grained information. Moreover, a mixed attention module is introduced into the network to extract more discriminating features. Finally, we use the DR-triplet loss to achieve more optimal directions of pair displacements to improve the retrieval performance. Experiments on two extended FG-SBIR datasets, QMUL-Shoe and QMUL-Chair, prove the effectiveness of our method. [ABSTRACT FROM AUTHOR]