The features of detected objects 104 determined at object detection 110 may be filtered according to one or more indexing criteria at indexing criteria filter 120, in some embodiments. For example, if object detection 110 performs face detection, then the extracted features may include face pose features such as pitch, yaw, and roll. In such a scenario, indexing criteria filter 120 may apply one or more threshold tests for different ones of the extracted features, such as pitch of a face pose between ?80° and 80°, a yaw of a face pose between ?90° and 90°, or a roll of a face pose between ?56° and 56.° In some embodiments, some extracted features may be common to many different types of detect objects. Brightness, sharpness, or confidence score for the output of the face detection, for example, may be such commonly extracted features. Thus, in some embodiments, indexing criteria filter 120 may evaluate extracted features for different types of detected objects with respect to a brightness greater than a minimum value, a sharpness greater than a minimum value, a confidence score for the output of the face detection greater than a minimum value, and/or various dimensions of a bounding box for the detected object.

In various embodiments, detected objects with features that do not satisfy the indexing criteria filter 120 may be excluded 106. Consider excluded object 156. Various feature values, such as brightness, bounding box size, or sharpness could have failed to exceed a minimum threshold value.