CVPR
2011
IEEE
13 years 8 months ago
2011
IEEE
Visual attributes expose human-defined semantics to object recognition models, but existing work largely restricts their influence to mid-level cues during classifier training....