Index filtering 320 may apply a filter corresponding to the type of object detection pipeline (e.g., a face filter for a face detection pipeline, a text detection filter for a text detection filter pipeline, and so on), in some embodiments. As noted earlier, in some embodiments, indexing criteria may not be linearly applied but may be weighted in different combinations. For example, for face detection pipelines alternative sets of criteria may be satisfied so that satisfying one of the criteria sets may allow the detected face to be included. For example, one criteria set may be satisfied by exceeding a minimum threshold of sharpness (e.g., 95%) and confidence (e.g., 95%), or a second criteria set may be satisfied by a pose with a pitch value, yaw value, and roll value within certain ranges, exceeding a minimum brightness value, exceeding a minimum sharpness value (e.g., which may be different than the other criteria set, such as >=40%), exceeding a minimum confidence value (e.g., which may be different than the other criteria set, such as >=80%), and a bounding box height and width greater than minimum values. In some embodiments, indexing filtering criteria could be staged so that a first pass filter may identify objects to definitively include (or exclude) whereas later stage indexing criteria could include performing further analysis of the detected object.