RAID: a relation-augmented image descriptor

摘要: As humans, we regularly interpret scenes based on how objects are related, rather than the themselves. For example, see a person riding an object X or plank bridging two objects. Current methods provide limited support to search for content such relations. We present raid, relation-augmented image descriptor that supports queries inter-region The key idea of our is encode region-to-region relations as spatial distribution point-to-region relationships between regions. raid allows sketch-based retrieval and requires minimal training data, thus making it suited even querying uncommon evaluate proposed by into large databases successfully extract non-trivial images demonstrating complex relations, which easily missed erroneously classified existing methods. assess robustness multiple datasets when region segmentation computed automatically very noisy.