局所的情報に基づく属性認識手法の検討

抄録

Attributes in various computer vision application are used as inter mediate representation. Especially in object recognition, attributes are defined as components of object, e.g. “ leg ” and “ furry ” in “ dog ” , “ stripe ” and “ V - Neck ” in “ T - Shirt ” . Thus localized information in image is effective to recognize attributes themselves . In this paper, we evaluated the influence of using localized information in training attributes.