Hierarchical ranking of facial attributes

摘要

In response to a query of discernable facial attributes, the locations of distinct and different facial regions are estimated from face image data, each relevant to different attributes. Different features are extracted from the estimated facial regions from database facial images, which are ranked in base layer rankings as a function of relevance of extracted features to attributes relevant to the estimated regions, and in second-layer rankings as a function of combinations of the base layer rankings and relevance of the extracted features to common ones of the attributes relevant to the estimated regions. The images are ranked in relevance to the query as a function of the second-layer rankings.

主权项

1. A method implemented by a computer comprising at least one processor for ranking images with respect to query attributes, the method comprising:
in response to a query comprising a plurality of specified facial attributes that are each discernable from face image data, estimating locations of a plurality of facial regions that are each relevant to different ones of the specified facial attributes, and wherein the plurality of facial regions are distinct and different from each other; extracting a plurality of different features from the estimated facial regions of each of a plurality of database facial images; ranking the database facial images in base layer rankings for different ones of the extracted features and the plurality of attributes as a function of relevance of the plurality of different features extracted from one of the images to each of single ones of the attributes relevant to one of the estimated regions and as a function of matching labeled rank-positions according to a base-layer training data set; ranking the database facial images in second-layer rankings as a function of bilinear combinations of the base layer rankings for a common ones of the attributes and as a function of matching labeled rank-positions according to a second-layer training data set; and ranking the images for relevance to the query as a function of respective combinations of the base layer rankings and the second-layer rankings for each of the different features extracted from the estimated facial regions of the plurality of database facial images.